Académique Documents
Professionnel Documents
Culture Documents
This book is a collection of fifteen essays that deal with issues at the
intersection of phenomenology, logic, and the philosophy of mathe-
matics. The first of the three parts, “Reason, Science, and Mathemat-
ics,” contains a general essay on Husserl’s conception of science and
logic, an essay on mathematics and transcendental phenomenology,
and an essay on phenomenology and modern pure geometry. Part II
is focused on Kurt Gödel’s interest in phenomenology. It explores
Gödel’s ideas and also some work of Quine, Penelope Maddy, and
Roger Penrose. Part III deals with elementary, constructive areas of
mathematics – areas of mathematics that are closer to their origins
in simple cognitive activities and in everyday experience. This part of
the book contains essays on intuitionism, Hermann Weyl, the notion
of constructive proof, Poincaré, and Frege.
i
P1: JZP
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 November 3, 2005 5:13
To my parents,
James D. and Beverly J. Tieszen
ii
P1: JZP
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 November 3, 2005 5:13
RICHARD TIESZEN
San Jose State University
iii
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo
Cambridge University Press has no responsibility for the persistence or accuracy of urls
for external or third-party internet websites referred to in this publication, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.
P1: JZP
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 November 3, 2005 5:13
Contents
v
P1: JZP
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 November 3, 2005 5:13
vi Contents
Bibliography 337
Index 349
P1: JZP
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 November 3, 2005 5:13
Acknowledgments
The essays collected here were written over a period of fifteen years. In
preparing them for publication in this volume I have modified them in a
few places, mostly for clarity and for continuity with other chapters in the
collection. I have also cut some material from a few of the essays. Some
overlap or repetition remains here and there, but the trade-off is that such
overlapping allows the essays to be read independently of one another.
In any case, I think that a little repetition is not onerous. It may even be
helpful to some readers. The Bibliography has been standardized, and in
the case of Husserl’s work in particular I have provided references to the
original publications, the relevant Husserliana editions, and the English
translations. Since Husserl’s works are usually composed in short sections
I have adopted the convention in the essays of referring to quotations
from Husserl’s works by providing the title of the work and the section
number. This method puts the reader in touch with the relevant texts but
allows for choice in consulting the different editions and languages. In
the case of Gödel’s writings I have followed the citation style used in Kurt
Gödel: Collected Works. The Bibliography for the present volume includes
very few works that were not cited in the original essays. I have included a
few new references where there has been some clear line of development
of an argument or point in the essays.
Chapter 1 was written for the volume Continental Philosophy of Science,
edited by Gary Gutting (Oxford: Blackwell, 2004). It appears here with
the permission of Blackwell Publishing.
Chapter 2 originally appeared under the title “Mathematics” in The
Cambridge Companion to Husserls, B. Smith and D. Smith (eds.) (Cambridge:
vii
P1: JZP
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 November 3, 2005 5:13
viii Acknowledgments
Acknowledgments ix
Over the years I have discussed ideas about phenomenology, logic, and
mathematics with many friends and colleagues. I have included the
original acknowledgments in the chapters themselves, but there have
been many other people who also deserve to be thanked for discussion,
comments, and suggestions.
My interest in the relationship of phenomenology to the exact sciences
goes back to my days in graduate school and college. During this period
I benefited most from interactions with Charles Parsons, Wilfried Sieg,
Howard Stein, Shaughan Levine, and Isaac Levi at Columbia University;
Robert Tragesser at Barnard; and J. N. Mohanty and Izchak Miller at the
Graduate Faculty of the New School for Social Research. Soon after I
arrived at Columbia, Charles Parsons gave a seminar on Husserl’s Logical
Investigations. This allowed me to deepen the study of Husserl’s works that
I had already begun at the New School and in college. At the New School,
J. N. Mohanty gave a year-long seminar on Husserl’s logical works. It was
this kind of in-depth seminar that made the Graduate Faculty unique.
As an undergraduate I benefited from studies of phenomenology with
Robert Welsh Jordan and of modal logic and other systems of logic with
Fred Johnson. It was during my time at Columbia that I met and began
the first of many discussions with Hao Wang about Gödel’s philosophical
views. Wang was one of the people I knew who most encouraged think-
ing about mathematics from a phenomenological standpoint. I also met
Dagfinn Føllesdal while I was a graduate student, and although he was
P1: JZP
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 November 3, 2005 5:13
x Acknowledgments
Introduction
Themes and Issues
2 Introduction
§1
In order to appreciate the approach I take to phenomenology in this
book it is necessary to realize that Husserl’s own thinking about logic and
mathematics went through several transformations. It will be useful to
situate my work in a general way with respect to three main stages that
can be discerned in Husserl’s writings on these subjects. Roughly speak-
ing, there is the early work of the Philosophy of Arithmetic (PA) and related
writings (1891–1900), the middle period of the Logical Investigations (LI)
(1900–1907), and the later period starting with the Ideas Pertaining to a
Pure Phenomenology and to a Phenomenological Philosophy (Ideas I) (1907–
1938). (For much more detail see Tieszen 2004.) Some of the central
changes that divide these periods from one another and that are relevant
to my work in this volume are as follows: the Philosophy of Arithmetic con-
sists of descriptive psychological and ‘logical’ investigations of arithmetic.
Husserl’s ontology at this point includes physical entities and processes
and mental entities and processes. The ‘ideal’ or abstract objects that are
clearly part of Husserl’s ontology from 1900 onward are not to be found
(in any obvious way) in PA. Husserl analyzed the origins of the natural
number concept in PA by focusing on mental processes of abstraction,
collection, and so on. As a result, some of the critics of the book, most
notably Frege, thought they detected a form of psychologism in PA. Psy-
chologism is the view that mathematics and logic are concerned with
mental entities and processes, and that these sciences are in some sense
branches of empirical psychology. Psychologism was a popular form of
naturalism about logic and mathematics in the late nineteenth century.
1 See, for example, Michael Dummett’s Origins of Analytic Philosophy, Michael Friedman’s A
Parting of the Ways: Carnap, Cassirer, and Heidegger and Reconsidering Logical Positivism, and
many of the writings of Dagfinn Føllesdal and J. N. Mohanty.
P1: IWV
052183782int.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 14:12
Introduction 3
4 Introduction
Introduction 5
§ 20). It is this ‘games meaning’ that takes center stage in strict formalist
views of mathematics. In order to illustrate the distinction, let us create a
miniature formal system right now. Suppose this formal system has only
two rules of inference:
Rule 1: If there are sign configurations of the form ∼ and then
derive the sign configuration .
Rule 2: If there is a sign configuration of the form ⊕ then derive the
sign configuration .
Suppose the alphabet of the formal system consists of the symbols P . . . Z,
P . . . Z , P . . . Z , the constants ∼, ⊕, and the signs (, ), and that we have
an appropriate definition of well-formed expressions.
Query: If we have the sign configurations P ∼ (Q ⊕ R) and P ⊕ S, then
is the sign configuration Q derivable? Yes. Apply Rule 2 to P ⊕ S to obtain
P. Apply Rule 1 to P and P ∼ (Q ⊕ R) to obtain Q ⊕ R. An application of
Rule 2 to Q ⊕ R will give us Q as output.
We could try many other queries and play with this little formal system
for a while. What, however, is this formal system about? Who knows, I just
concocted it. It need not be about anything. In order to derive Q from
the sign configurations that are given it is not necessary or even useful to
know the meaning of the expression, if any, for which Q has been chosen.
It is as though we are playing by the rules of a particular game and we
need not be concerned with what the signs are about.
One can create indefinitely many formal systems at will. In the space
of all possible formal systems, however, we find that some formal systems
are correlated with existing parts of mathematics, are meaningful to us,
interesting, or useful. If mathematics consists only of formal systems, then
how could we single out those systems that actually correlate with existing
parts of mathematics, that are meaningful, interesting, or useful?
Consider another example. Suppose I construct an axiomatic formal
theory in the language of first-order logic with identity and function
symbols. One can use standard rules of inference that are associated with
such a theory. The theory has three axioms:
Axiom 1: (∀x)(∀y)(∀z) g(x, g(y, z)) = g(g(x, y), z)
Axiom 2: (∀x) g(x, a) = x
Axiom 3: (∀x) g(x, fx) = a
Now would a sign configuration such as
(∀x)(∀y)(∀z)(g(x, z) = g(y, z) → x = y)
P1: IWV
052183782int.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 14:12
6 Introduction
Introduction 7
are also some themes (e.g., the concern for the ‘origins’ of concepts of
logic and mathematics) that remain more or less constant throughout
the stages. In the essays that follow I am always writing from the point
of view of the later two stages in Husserl’s work, even when I write about
Husserl’s earliest work. It will not be possible to understand what I am
doing in the chapters unless this is kept in mind. Husserl himself did not
have the opportunity to rewrite such works as the Philosophy of Arithmetic
from the perspective of his more mature philosophy. It is left to his readers
to try to rethink the early work on the philosophy of arithmetic and the
philosophy of geometry in terms of the later ideas and methods.
My earlier book, Mathematical Intuition: Phenomenology and Mathematical
Knowledge, was also written from the perspective of ideas in transcendental
phenomenology. The book was focused on mathematical intuition in the
case of natural numbers and finite sets, and the idea was to see what kind
of account one could develop in this case, on the basis of some of the
recent literature on mathematical intuition but also on Husserl’s later
ideas on intentionality, meaning, ideal objects, possibilities of dynamic
fulfillment of mathematical intentions, the analysis of the origins of math-
ematical concepts in everyday experience, and related views about acts
of abstraction, reflection, formalization, and so on. Some of the essays in
the present volume, especially Chapters 11 through 15, need to be under-
stood from this perspective. I am concerned with the matter of how far we
can push the idea of the fulfillment or fulfillability of intentions that are
directed toward particular mathematical objects. What is the best way, for
example, to understand talk about the (potential) presence or absence
to human consciousness of particular natural numbers? Are there limits
on the intuition of particular mathematical objects? This can be seen as
an effort to understand the relationship between the more intuitive parts
of mathematics that have their origins in everyday experience (for exam-
ple, arithmetic and elementary geometry) and the more conceptual and
rarefied parts of mathematics (such as higher set theory) where it appears
that we cannot have complete or fully determinate intuitions of particular
mathematical objects (even if we can engage in ‘objective’, meaningful,
eidetic thinking that appears to be directed toward such objects).
Because I have compared some of Husserl’s ideas in the elementary
parts of mathematics that have their origins in everyday experience with
ideas in intuitionism (see Part III) one might form the impression that I
think Husserl himself was an intuitionist. As should be obvious in many
of the following chapters, this would amount to a misunderstanding of
what I am doing. Husserl was not an intuitionist in the style of Brouwer or
P1: IWV
052183782int.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 14:12
8 Introduction
Heyting. There are many differences between his views and those of the
traditional intuitionists. One could list the differences. One of the central
differences, for example, is that from 1900 onward Husserl has ideal ob-
jects such as meanings in his ontology and he speaks about these objects
in a way that is more platonistic than would be pleasing to Brouwer or
Heyting. Mental entities and processes are required for knowing about
natural numbers, for example, but the objects known about – natural
numbers – are not themselves mental entities. Rather, they are ideal ob-
jects. There are also some important differences concerning solipsism,
the explicit recognition of the intentionality of human consciousness,
meaning theory, the place of formalization, the views of what can be in-
tuited, and the like. Husserl holds, for example, that there is intuition of
meanings or of essences, and of the relations of essences to one another,
but it is not clear that traditional intuitionism would have a place for this
kind of intuition.
At the same time, it is very interesting to compare Husserl’s explo-
rations of logic and mathematics with those of the intuitionists. Husserl
never tired of arguing that in order to do justice to logic and mathematics
we must investigate the subjective side as well as the objective side of these
sciences. We need to consider not only the ideal objects and truths of logic
and mathematics but also the subjective acts and processes by virtue of
which we come to know about objects and truths. Such acts and processes
include carrying out sequences of acts in time, abstracting, collecting, re-
flecting, and various forms of memory and imagination. No one in recent
times has done more to investigate the subjective side of logic and mathe-
matics than Husserl and people such as Brouwer and Heyting. There are
bound to be interesting and important points of contact. Some of these
points of contact were in fact already being explored in Husserl’s time
by such individuals as Oskar Becker (who was one of Husserl’s research
assistants), Hermann Weyl (who was influenced in some ways by Husserl),
Felix Kaufmann (also influenced by Husserl), and Heyting himself. Weyl,
for example, comes close to identifying the phenomenology of mathe-
matics with intuitionism in some of his comments. As I have indicated, I
think this identification goes too far. It does seem to me, however, that
there are ideas about subjectivity, intersubjectivity, meaning, intuition,
internal time, and other topics in transcendental phenomenology that
can be used to support and develop some ideas in intuitionism. Indeed,
this is the tack I take in some of the chapters in Part III. In any case, I
think it is not a good idea to act as if this period in the development of
phenomenological ideas about mathematics did not exist.
P1: IWV
052183782int.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 14:12
Introduction 9
10 Introduction
§2
It will be useful to include some specific comments on each of the chap-
ters in this collection in order to put them into proper perspective and to
indicate some connections between them. The essays are grouped into
three categories. Part I of the volume contains three essays on reason, sci-
ence, and mathematics from the viewpoint of phenomenology. The first
chapter provides an overview of Husserl’s claims about reason and about
logic as a theory of science. It takes up the idea of logic as the science
of all possible sciences (as mathesis universalis), including Husserl’s ma-
ture view (in Formal and Transcendental Logic) of the three levels of what
he calls ‘objective formal logic’, along with some of his ideas on man-
ifold theory and formal ontology. The essay follows Husserl’s thinking
about science through his late work on the crisis of the modern sciences.
That crisis stems from misapplications of various forms of naturalism
and objectivism. The scientism that comes in for criticism in Husserl’s
late work is a view that has abandoned philosophical rationalism. Just
as there can be science within reason, there can also be science with-
out reason. The second chapter in Part I focuses on some problems in
the philosophy of mathematics in particular and on ways they might be
approached from the perspective of transcendental phenomenology. It
should be read as a general overview that is filled out in various ways in
subsequent chapters. The third essay focuses in more detail on geometry.
It discusses Husserlian views about ‘ideation’ or the intuition of essences
in connection with modern pure geometry, the idea that there are dif-
ferent formal systems of geometry and different spatial ontologies, and
some Husserlian themes about the origins of geometry. It considers the
idea of creating new variants of geometry based on the formalization of
Euclidean geometry. The essay also discusses the concern for the consis-
tency of the resulting systems, especially when we have left behind the
more familiar and intuitive domains of two and three dimensions.
Part II of the volume is centered on a series of essays on Kurt Gödel’s
interest in Husserl’s phenomenology. Gödel had turned his attention
P1: IWV
052183782int.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 14:12
Introduction 11
12 Introduction
Introduction 13
that the expression “round rectangle” has a meaning. There is, however, a
“dissonant unity” of the intended essences “round” and “rectangle” here.
These essences are incompatible. One can of course also reflect upon
and try to clarify the meaning of instances of expressions of the form
“P is essential to x.” It might be helpful to keep these points in mind if
one wants to read Gödel’s ideas on phenomenology in connection with
Husserl’s own views.
Finally, I would like to note that although there are many interest-
ing connections between Gödel and Husserl, it might be the case that
Gödel’s platonism differs in some respects from what we find in Husserl.
Gödel evidently favored Husserl’s transcendental phenomenology, and
it seems to be possible to make fairly robust platonistic claims from this
perspective. We can evidently say, for example, that mathematical objects
are mind independent and unchanging, but now we always add that they
are constituted in consciousness in this manner, or that they are consti-
tuted by consciousness as having this sense (meaning). One can evidently
say that it is not necessary for the existence of mathematical objects or
truths that anyone ever be aware of them or that there ever be expressions
for them, but we now add that they are constituted as having this sense.
They are constituted in consciousness, nonarbitrarily, in such a way that
it is unnecessary to their existence that there be expressions for them
or that there ever be awareness of them. Gödel would perhaps be happy
with such a view, but I can imagine objections to it as a reading of either
Gödel or Husserl. The matter certainly deserves further investigation.
Another issue is that modern, transfinite set theory of the sort that Gödel
was concerned with in some of his work raises problems that were not
directly addressed by Husserl. Should we really accept the existence of
the huge, impredicatively specified transfinite sets that we speak about in
higher set theory? Can we legitimately speak about the constitution of the
being of such sets? What about the sets postulated by some of the axioms
that have been offered in order to extend Zermelo-Fraenkel set theory?
Some of Gödel’s comments indicate that he was prepared to accept the
full ontology of Zermelo-Fraenkel set theory (and more), but it is possible
that Husserl would not go so far. In any case, I think that according to
Husserl’s meaning theory these parts of mathematics would indeed count
as meaningful. It is just that they demand a more penetrating clarifica-
tion of meaning than some other parts of mathematics. Gödel evidently
thought that there was a sharp distinction between Husserl’s meaning
theory and other theories of meaning that were on offer but that were
antithetical to his purposes.
P1: IWV
052183782int.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 14:12
14 Introduction
Introduction 15
To say explicitly that exactly two things have the property X we adopt the
following characterization:
Similarly,
(∀X)(0X ↔ ¬(∃y)Xy),
(∀X)(1X ↔ (∃x)(∀y)(Xy ↔ y = x)),
(∀X)(3X ↔ (∃x)(∃y)(∃z)((x = y ∧ y = z ∧ x = z)∧
(∀w)(Xw ↔ (w = x ∨ w = y ∨ w = z)))),
and so on.
Now consider a statement such as 2 + 2 = 4. On the kind of analysis
that Frege gives this goes over into a statement of higher-order logic of
the form
To see the point of Husserl’s claim (and compare this with Poincaré’s
remark in § 3 of Chapter 14) we can use a substitution test. Suppose a
person P knows that 2 + 2 = 4. Does it follow with necessity that
16 Introduction
Introduction 17
I believe that presenting these essays in one volume will allow the
reader to see many connections that would be missed if only a few of
the essays were read or if they were read in isolation. The essays can be
read independently, but they are probably most profitably read in con-
nection with one another. It is my hope that the book will help to carry
forward ideas in the phenomenological tradition about logic and math-
ematics. It contains new perspectives on and new ideas about a variety of
topics in the philosophy of logic and mathematics: the role of intention-
ality in mathematics, mathematical intuition, platonism, constructivism,
impredicativity, logicism, varieties of formalism, nominalism, minds and
machines, and the like.
P1: IWV
052183782int.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 14:12
18
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
part i
REASON, SCIENCE, AND MATHEMATICS
19
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
20
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
The reason for the failure of rational culture, as we said, lies not in the
essence of rationalism itself but solely in its being rendered superficial, in
its entanglement in ‘naturalism’ and ‘objectivism’.
(Husserl, “The Vienna Lecture,” 1935)
In preparing the final version of this chapter I benefited from the comments of the many
participants in the Science and Continental Philosophy conference organized by Gary
Gutting and held at the University of Notre Dame in September 2002. I especially thank
Ernan McMullin, Gary Gutting, Karl Ameriks, Michael Friedman, Philip Quinn, Hans-Jörg
Rheinberger, Terry Pinkard, and Simon Critchley.
21
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
of how modern scientific culture has fallen away from its higher calling. I
will discuss each of these aspects of Husserl’s philosophical thinking. (For
some earlier studies of these topics see, e.g., Gurwitsch 1974; Heelan 1989;
Ströker 1988, 1987a, 1979.)
The critique of the modern sciences in Husserl’s later work did not
spring forth ex nihilo. There are seeds of it in his earliest work. The
Philosophie der Arithmetik (PA), for example, is already premised on the
view that only the philosopher can provide the kind of deeper reflection
on arithmetic that the mathematical technician either cannot or will
not provide. In this first book we are presented with an analysis of how
this science is founded on the everyday experience of groups of sensory
objects, how abstraction from this basis and a kind of formalization are
required to get the science of arithmetic off the ground. Arithmetic is
built up from ‘higher’ cognitive activities that are founded on ordinary
perceptual acts. The idea of analyzing ‘origins’ of concepts, which Husserl
inherited from some of his teachers (especially Brentano), is already at
work here, and Husserl will continue to insist throughout his career on
the value of this kind of analysis. Ideas of this type will figure into his
later charge that the modern sciences have forgotten their origins in the
everyday practices of the lifeworld.
By the time of the Logical Investigations (LI) Husserl tells us that philoso-
phers have the right, indeed the duty, to examine critically the founda-
tions of the sciences. Philosophy and, in particular, phenomenology are
needed to supplement the sciences. After 1907 or so phenomenology is
portrayed in ever more detail as the transcendental and a priori ‘science’
that investigates the essence of the sciences (as well as other domains of
human experience). Scientists themselves are technicians who build up
theories and methods without insight into the essence of these theories
and methods or into the conditions for their possibility. Scientists are
concerned more with practical results and mastery than with essential
insight. For just this reason, the sciences are in need of continual episte-
mological reflection and critique of a sort that only the philosopher can
provide. Scientists are oriented toward their objects but not toward the
scientific thinking itself in which the objects are given. It is the phenome-
nologist who will study the essential features of this thinking. Husserl
pictures the work of the philosopher and the scientist as mutually com-
plementary (Husserl LI, “Prolegomena to Pure Logic,” § 71). The philoso-
pher does what the scientist cannot and will not do if she is to practice
her science, and the scientist does what the philosopher cannot do qua
philosopher.
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
outside time altogether but holds that they exist at all possible times. As
one would expect, ideal objects are also acausal. Thus, we seem to be pre-
sented with a rather platonistic view of ideal objects, but this ‘platonism’
becomes rather nuanced after Husserl takes his turn into transcenden-
tal idealism around 1908. In these later writings Husserl speaks of how
the sense of every existent is constituted in the subjectivity of conscious-
ness. This means that the sense of ideal objects and truths as transcen-
dent, nonmental (and, hence, in a sense, as mind independent), partially
given, omnitemporal, and acausal is itself constituted in the subjectivity of
consciousness (see, e.g., Husserl’s Formal and Transcendental Logic (FTL),
§ 94).
Pure logic is concerned with ideal meanings, and it is judgments that
express these ideal meanings. Judgments refer to objects or states of af-
fairs by way of their meanings. The meanings expressed by judgments
are ideal and the objects to which we are referred by judgments may
themselves be either real or ideal. The objects of pure logic and pure
mathematics are ideal. Judgments have a form and a ‘matter’ (or con-
tent). Two judgments with different ‘matters’, for example, may have the
same form: ‘This house is red’ and ‘This table is blue’ both have the form
‘This S is P’ (this is only a partial formalization). Among other things, pure
logic will therefore need to track features of judgments and other types
of expressions, the ideal meanings expressed by judgments, the objects
referred to, and the form and matter in each case. Husserl’s view of all
of this, inaugurated in the LI and developed up through FTL, is mapped
out in his stratification of ‘objective formal logic’ into three levels. I will
give a brief overview of the conditions for the possibility of science that
are included in this threefold stratification.
A fundamental condition for the possibility of any science is that it op-
erate with judgments that not only are formally well formed but express
unified or coherent meanings. Thus, at the bottom or first level of any
possible science we have a priori or ‘universal’ grammar. This is not to
be a psychologistic or anthropological science. Husserl thinks there will
be a priori rules of grammar for both the form and the matter of ex-
pressions. The basic idea at this level is to lay out the rules and methods
for determining whether or not a string of signs (words) is meaningful.
This will require purely formal grammar, in order to distinguish formally
well-formed strings of signs from strings that are not well formed. Here
Husserl seems to have in mind what we now think of in laying out the
grammar of purely formal languages. Using one of his own examples, we
could say that ‘This S is P’ meets the relevant formal conditions, whereas
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
a string such as ‘This is or’ does not. Husserl distinguishes simple from
complex meanings and he wants to know the rules for forming judgments
as meaningful wholes from meaningful parts. Already in Investigation IV
of his Logical Investigations he had applied the theory of parts and wholes
from Investigation III to this question and concluded that there must
be a priori laws of grammar. What is needed, in addition to a purely
formal grammar, is a theory of semantic categories to determine which
substitutions of matter in the forms will give us something meaningful
and which will give us mere nonsense (Unsinn). Consider, for example,
the following two substitutions for ‘This S is P’: ‘This tree is green’ and
‘This careless is green’. The former is a judgment. It expresses a unified
meaning. The latter, however, is simply nonsense. Each part of it is mean-
ingful but the whole formed from the parts is not. An expression from
the wrong category has been substituted in place of ‘S’. Roughly speak-
ing, a categorial grammar would allow nominal material to be replaced
by nominal material, adjectival material by adjectival material, relational
material by relational material, and so on. Adjectival material could not
be freely replaced by nominal material, and so on. Husserl allows that
false, foolish, or silly judgments may result from the substitutions permit-
ted but they would still be meaningful judgments. We might, for example,
obtain a judgment such as ‘This blue raven is green’. This is not meaning-
less (nonsense), but it is false. Indeed, it is an a priori material absurdity
or inconsistency, on the assumption that nothing that is blue all over can
be green. On the other hand, “This careless is green” is nonsense.
Level two of ‘objective formal logic’ is the ‘logic of noncontradiction’
or ‘consistency logic’. The strings of words admitted at level one express
unified meanings. The next question we can ask about such (sets of)
judgments is whether or not they are consistent. Here we also need to
track formal and material versions of this question. A form such as ‘There
is an S and there is no S’ is formally contradictory. It is a purely formal
a priori absurdity. There may also be judgments that are not formally
contradictory but that are instead materially inconsistent, for example,
‘This blue raven is green’. The latter expression is meaningful but Husserl
says it is countersensical (Widersinn). It is a material a priori absurdity.
Thus, there are both formal and material a priori absurdities. A judgment
such as ‘This raven is orange’, on the other hand, is meaningful and is
not a material a priori absurdity. It is a synthetic a posteriori judgment,
the kind of judgment whose truth or falsity requires sense experience.
Husserl in fact distinguishes synthetic a priori from synthetic a poste-
riori judgments along these lines (see Husserl LI, Investigation III, and
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
Husserl Ideas I, §§ 10 and 16). Both require material concepts but the
former depend on purely a priori relations among such concepts. For
example, if an object x is blue all over, then it cannot be green. If an ob-
ject x is red, then x is spatially extended, but the converse does not hold.
If an object x is red, then it is colored, but again the converse does not
hold. Sense experience could not tell against such truths. What Husserl
calls ‘regional ontologies’ (discussed later) are made up of synthetic a
priori truths for different domains of cognition. Analytic a priori truths,
on the other hand, are purely formal. They depend on purely formal
a priori laws. Husserl thus speaks of purely formal logic as ‘apophantic
analytics’. (Apophansis is the Greek term for judgment or proposition.)
The ontological correlate of purely formal logic is formal ontology. In this
connection, Husserl also distinguishes formalization from generalization
or, if you like, formal from material abstraction. Abstraction of the form
from a material proposition is different from the kind of abstraction in-
volved in moving from species to genus. The relationship of the form of a
proposition to its material instances does not require concepts with con-
tent in the same way that these are required in setting out genus/species
relationships. For example, a concept lower in a species/genus hierarchy
(e.g., being red) implies all of the concepts (e.g., being extended) above
it (but not the converse). It is for reasons of this type that there can be
synthetic a priori judgments and regional ontologies.
At the second level of logic we should aim to distinguish judgments
that are countersensical from those that are not. If a judgment is coun-
tersensical, it is meaningful, but it is not possible that there are objects
corresponding to it. There could be no state of affairs that corresponded
to it. On the other hand, if it is consistent, then it can have corresponding
objects or states of affairs.
Husserl develops at this level the idea of purely formal ‘apophantic’
logic(s). Apophantic logic would just be the logic of forms of judgments,
broadly conceived. It would presumably include forms of any kinds of
meaningful judgments. One might speak of different logics here, but
Husserl seems to have in mind some unified conception of these in one
overarching logic. Consistent apophantic logic(s) would stand as a con-
dition for the possibility of any science since we cannot have sciences that
contain formally contradictory judgments.
Apophantic logic is conceived of in terms of axiomatic formal systems.
Husserl sometimes refers to this level of consistency logic as ‘consequence
logic’. We view logic as an axiomatic formal system in which one derives
consequences of the axioms on the basis of formal rules of inference.
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
The idea that judgments express meanings by virtue of which they refer
to objects or states of affairs is mirrored at this second level of logic, in
the following sense: if a set of judgments (including a formal system)
is consistent, then objects or states of affairs corresponding to the set
of judgments are possible. Husserl calls the ‘ontological correlate’ of a
consistent formal axiomatic system a manifold. Since it is the correlate
of the mere forms of judgments it constitutes a purely formal ontology.
In apophantic logic we are focused on the forms of judgments and the
features of judgments as expressions of meanings, whereas in formal
ontology we are focused on the possible objects or states of affairs re-
ferred to by such judgments, solely with respect to their form. In formal
ontology we are concerned with the most general and formal notions
of object, state of affairs, property, relation, whole, part, number, set,
and so on. In the particular sciences these notions would be ‘material-
ized’ or specified in different ways. Geologists, for example, would speak
about particular kinds of objects (e.g., rocks, mountains) and particu-
lar kinds of parts, properties, and relations of these objects. Physicists
would specify their objects and properties differently than geologists or
biologists but would still be using the notions of object, part, property,
and relation. In the different sciences we would have different regional
ontologies. (For studies of the natural sciences in particular see Hardy
and Embree 1993; Heelan 1983; Kockelmans 1993; and Kockelmans and
Kisiel 1970.)
We can ask not only whether formal systems are consistent but also
whether they are complete. Husserl introduced the notion of a ‘definite’
formal system early on in this thinking, and it seems to be the notion
of a formal system that is both consistent and complete. He frequently
mentions how he and Hilbert arrived at such a notion independently of
one another (see, e.g., Husserl Ideas I, § 72). Husserl also introduces the
notion of a definite manifold as the ontological correlate of a definite
formal system. It is the definite purely formal ‘world’ that corresponds to
a definite axiomatic formal system.
At the third level we have what Husserl calls ‘truth logic’ (Wahrheits-
logic). At this level we are concerned with the truth or falsity of judgments.
If judgments are consistent, then it is possible that there are objects corre-
sponding to the judgments and it is possible that the judgments are true.
For judgments to be considered true, however, it appears that we need
more than mere consistency. Truth or falsity is determined by intuition,
or by what Husserl calls ‘meaning-fulfillment’. Intuition takes place in
sequences of acts carried out through time, and it provides evidence that
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
one could begin to speak about what is in principle fulfillable and what
is not. Natural numbers themselves were now clearly taken to be ideal
objects. In any case, the distinction between the intuitive and the merely
symbolic was always present in Husserl’s philosophy of science. Some
parts of mathematics, logic, and even natural science would evidently al-
ways have to be considered merely symbolic or conceptual. (We might
keep the symbolic and the conceptual separate here, reserving the former
term for the merely algorithmic ‘games-meaning’ of signs and the latter
term for the full meaning-intention of signs.)
Within this architectonic Husserl, as we said, distinguishes formal on-
tology from various regional ontologies. Regional ontologies are a priori
ontologies of particular regions of being. When forms are filled in with
‘material’ or content, we are considering specific kinds of objects, proper-
ties, relations, parts, wholes, sets, and so on. All of the different specific sci-
entific theories would have different regional ontologies. Transcendental
phenomenology itself is a regional science. It is the science of conscious-
ness with all of its various structures and characteristics. Each science
has its own objects, with their own properties, and so forth. Husserl is
generally an antireductionist about the specific sciences. Part of the rea-
son for this derives from his view that judgments in the various sciences
express their own meaning-intentions, and it is by virtue of these meaning-
intentions that we are directed toward objects. Reductionist schemes may
very well fail to respect differences in meaning-intention, along with the
different implications and purposes reflected in these. For certain sci-
entific purposes they might, if feasible, be useful. Eliminative reduction-
ism in particular might diminish the many aspects or perspectives under
which the world could be viewed and investigated. Eliminativism could
hinder scientific work and perhaps even be dangerous, blinding us to
important phenomena.
One very broad distinction among the sciences, as has been suggested,
is that some are eidetic or a priori and some are empirical. Formal and
regional ontologies are a priori, dealing, respectively, with the formal (an-
alytic) and material (synthetic) a priori. Husserl also says that among the
eidetic sciences some are exact and some are inexact. Mathematics and
logic are exact. They trade in exact essences and exact objects. Among
the inexact eidetic sciences Husserl counts transcendental phenomenol-
ogy. It is, Husserl says, a ‘descriptive’ science that deals with inexact or
‘morphological’ essences, for example, the essence ‘consciousness’. Un-
like empirical psychology, however, it is not a causal-explanatory science.
It does not seek causal generalizations. Underneath this rationalistic
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
This is really only part of a larger whole, however, that includes human
subjects as those who are doing the judging, thinking, remembering, and
so on. The whole picture is more like this:
The positive or ‘objective’ sciences deal with only part of a larger whole,
and a dependent part at that. Husserl calls the broad conception of logic
or of science in which we do not omit the role of the human subject a
‘transcendental logic’. As in Kant’s conception, it is a ‘logic’ or ‘science’ in
which we do not abstract from the possible experience of human beings.
The expression transcendental logic is in fact often used as another ex-
pression for transcendental phenomenology itself. Transcendental phe-
nomenology is to be the science of the subjective and intersubjective side
of experience, of consciousness and its object-directedness in any domain
of conscious experience. It is to be the science of the essential features
and structures of consciousness that provides the philosophical founda-
tion of the sciences. If we are really interested in the conditions for the
possibility of science, we cannot forget that, at bottom, it is the human
subject who makes science possible. It is human subjects who bring about
or constitute the sciences over time and who hand down the sciences from
generation to generation.
On Husserl’s analysis there are different types and levels of conscious-
ness. Science is built up from the lifeworld experience of human sub-
jects on the basis of acts of abstraction, idealization, reflection, formaliza-
tion, and so on. The most basic, founding experiences are the everyday
lifeworld activities, practices, and perceptions of people. The lifeworld,
Husserl says, is “the intuitive surrounding world, pregiven as existing for
all in common” (Husserl The Crisis of the European Sciences and Transcen-
dental Phenomenology (Crisis), §§ 33–34). All of our activities, including our
loftiest sciences, presuppose the everyday practical and situational truths
of the lifeworld. Our praxis and our prescientifc knowledge in the life-
world play a constant role in all of our activities. The lifeworld was there
for us before science, and even now human beings do not always have
scientific interests. We have the intuited, everyday world that is prior to
theory and then the various theories that are built up from this basis.
Science thus presupposes the lifeworld as its starting point and cannot
therefore replace this world or substitute something else for it. Human
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
subjects are the meaning-givers and interpreters who produce the sci-
ences from this basis and who can choose to be responsible about them
or not.
Various fields of ‘objective’ science must therefore be correlated with
the subjective if we are to do justice to the sciences. To see things whole
we must deal with both objectivity and subjectivity. Otherwise science
is naı̈ve and one-sided. We would have only the positive, objective sci-
ences in which the human subject is not remembered in the scientist’s
work. The human subject we are talking about here, moreover, should
not be the reduced, dessicated subject that is presented to us by the
objective sciences. It should not be the subject as interpreted through
these sciences, which are themselves already founded on more basic
forms of human experience. As we said, the positive ‘objective’ sciences
are not foundational but are themselves founded on our lifeworld ex-
perience. In this manner, Husserl turns the table on those who think
of the sciences and technology as providing the fundamental ways of
knowing, understanding, and being, or who would value the positive
sciences and technology above all else. It is this turn of thought, this cri-
tique of scientism, that has resonated with so many Continental philoso-
phers for so long, from existentialists up through various postmodern
philosophers.
Husserl was arguing long before such philosophers as Thomas Nagel
that the positive sciences cannot in principle understand or do justice
to the human subject (see, e.g., “The Vienna Lecture” included with
Husserl’s Crisis). They are ‘objective’ sciences that make of the human
subject a kind of scientific object, an objectified subject. Husserl’s view
thus has many implications for psychology and the social or human sci-
ences, as can be seen in the secondary literature on the subject (see, e.g.,
Natanson 1973).
In order to obtain the objective sciences one abstracts away from many
features of experience. To abstract and to idealize are automatically to
simplify, leaving behind some of the complexity and richness of concrete
experience. Of course in the sciences such simplifications and idealiza-
tions help us to get a grip on things and make the work manageable.
The simplification serves a purpose and might sometimes be put to good
use, but one should not forget that it does not give us the complete
picture. In the objective sciences one abstracts, for example, from the
qualitative features of experience to focus on quantitative features. Cal-
culation and calculational techniques come to the fore. Primary qualities
of phenomena are highlighted and secondary qualities are marginalized.
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
of this, but it is quite different from being directed toward the objects to
which the formula is being applied. The objects are now the signs that
are being manipulated in the calculation. In this manner there can be a
complete displacement of concern (see Tieszen 1997b).
The skills and abilities associated with calculation and other forms of
technical work will obviously be valued all the more in this kind of envi-
ronment. This in turn fosters specialization in and professionalization of
domains of scientific work. Technical knowledge and pragmatic success
take center stage while other forms of knowledge and understanding tend
to be marginalized. A pragmatic instrumentalism is the natural outcome
of this shift in values and goals. This might be a pragmatic instrumental-
ism in which efficiency and the control and domination of nature are first
principles, an instrumentalism that tends to view nature and everything
in it as resources or ‘input’ for scientific/technological processes.
There is clearly potential here for a kind of alienation from reality, a
distancing from the basic foundations of knowledge and understanding.
This is an alienation made even worse by forgetting about these origins. If
we think of Husserl’s slogan “Back to the things themselves,” then it is an
alienation from ‘the things themselves’. One might say that the possibility
of such alienation attends the possibility of inauthentic ways of thinking
and being, as opposed to more authentic ways. All of this arises with the
abstractions and idealizations that make science possible. It arises from
taking something as an independent whole that is really only a dependent
part of a larger whole. It is to recede from a more holistic perspective.
In the modern natural sciences there is, Husserl thus says, a surrepti-
tious substitution of the mathematically structured world of idealities for
the only real world, the one that is actually given through perception: the
everyday lifeworld. This substitution already occured as early as Galileo,
and it was subsequently passed down through the generations (Husserl
Crisis, § 9h). What has happened is that the lifeworld, which is the foun-
dation of the meaning of natural science, has been forgotten. A type of
naiveté has developed. Galileo is at once both a discovering and a con-
cealing genius.
Although the techniques and methods of the sciences are handed
down through the generations their true meanings, as we have said, are
not necessarily handed down with them. It is the business of the philoso-
pher and phenomenologist to inquire back into the original meanings
through an eidetic analysis of the sedimentation involved. There is a ‘his-
torical meaning’ associated with the formations of the sciences, but, as
mentioned earlier, Husserl is not interested primarily in empirical history.
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
He is interested in finding the a priori unity that runs through all of the
different phases of the historical becoming and the teleology of philoso-
phy and the sciences (Husserl Crisis, § 15). Husserl says that it is a ruling
dogma that there is a separation in principle between epistemological
elucidation and historical explanation, or between epistemological and
genetic origin. Epistemology cannot, however, be separated in this way
from genetic analysis. To know something is to be aware of its historicity,
if only implicitly. Every effort at explication and clarification is noth-
ing other than a kind of historical disclosure. The whole of the cultural
present implies the whole of the cultural past in a contentually undeter-
mined but structurally determined generality. It implies a continuity of
pasts that imply one another. This whole continuity is a unity of tradi-
tionalization up to the present. Here Husserl speaks about unity across
difference on a global scale, not just in the case of the unities through
difference that arise for us in our own personal cognitive life. Of course
in the latter case too there is a temporality and a ‘history’. Anything his-
torical has an inner structure of meaning. There is an immense structural
a priori to history.
Historicism is the view that there could be no such historical a priori,
no supertemporal validity. It claims that every people has its own world.
Every people has its own logic. Husserl responds by pointing out some
of the background asumptions that are necessary for factual historical
investigation to occur at all (see Husserl Crisis, Appendix VI). These are
what we must know or assume before we can even get started with any
factual historical investigation. In spite of all the indeterminacy in the
horizon of ‘history’, it is through this concept or intention that we make
our historical investigations. This is a presupposition of all determinabil-
ity. But what kind of method can we use to make apparent to ourselves the
universal and a priori features? We need to use the method of free varia-
tion in imagination in which we run through the conceivable possibilities
for the lifeworld. In this way we remove all bonds to the factually valid
historical world. We determine what is necessary and invariant through
all of the contingencies and variations.
In the case of Euclidean geometry, for example, Husserl says that only
if the necessary and most general content (invariant through all con-
ceivable variation) of the spatiotemporal sphere of shapes is taken into
account can an ideal construction arise that can be understood for all
future time and thus be capable of being handed down and reproduced
with an identical intersubjective meaning. Were the thinking scientist to
introduce something time bound into her thinking, something bound
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
to what is merely factual about her present, her construction would like-
wise have a merely time-bound validity or meaning. This meaning would
be understandable only to those who shared the same merely factual
presuppositions of understanding. Geometry as we know it would there-
fore not be possible.
Husserl thus argues against a relativism that would deny the ideal, om-
nitemporal character of sciences such as geometry, arithmetic, and logic,
but he also argues against any absolutism that would cut off the truths of
such sciences from their relation to human subjects. The objective and
the subjective can only be properly understood in relation to one an-
other, each conditioned by the other. One can say the same thing about
the real and the ideal.
The things taken for granted in the positive sciences should, according
to phenomenological philosophy, be viewed as ‘prejudices’. As ideas and
methods become sedimented over the years they become, quite literally,
part of a prejudged mass of conditions and assumptions. In this manner
various obscurities arise out of a sedimentation of tradition. One should
subject such prejudgments again and again to critical, rational judgment.
The genetic investigation of phenomenology is thus meant to allow us
to become aware of such prejudices and to enable us to free ourselves of
various presuppositions. It is therefore supposed to be the deepest kind
of self-reflection aimed at self-understanding. Husserl sets up such pre-
suppositionlessness as an infinite, regulative idea that we may never reach
but that we should nonetheless never abandon. Husserlian hermeneutics
would thus be rather different in some ways from that, for example, of
Heidegger or Gadamer.
§ 4 Conclusion
These kinds of reflections lead to a host of issues that have distinguished
Continental thinking about science from the analytic philosophy of sci-
ence. Much of the analytic philosophy of science has tended to focus
on questions that are more internal to the scientific enterprise. There
have been periods during which some of the important questions raised
by Husserl’s critique of the sciences would simply never have occurred
to those working in the analytic philosophy of science. Consider those
circumstances, for example, in which the historical dimensions of sci-
ence have been overlooked or devalued or in which the hypotheses
and explanatory schemes of the positive sciences have been viewed as
nonhermeneutical or value free. Husserl’s work, by way of contrast, opens
P1: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:13
onto many issues involving the broader social, political, historical, and
ethical dimensions of science and technology.
Husserl can be read as providing a corrective to scientism, to the view
that it is only the positive sciences that supply knowledge, understanding,
and truth. He is not arguing that the positive sciences do not have their
place or do not have any value. Rather, we need to see them in their
proper perspective. We need the correct balance. If scientism is extreme
in one direction, then a view that completely rejects a place for the positive
sciences is too extreme in the other direction. Scientism can be a kind
of irrationalism and it can in its own way be blind, but there can also
be an antiscientific kind of irrationalism and blindness. It seems to me
that neither form of irrationalism would be acceptable to Husserl. It is
Husserl’s emphasis on reason that distinguishes him both from many
analytic philosophers and from many Continental philosophers.
The positive, objective sciences both reveal and conceal. Phenomeno-
logical philosophy, aiming toward science in a broader sense, tells us to
retain what is revealed by the positive sciences (subject to critical scrutiny,
responsibility, and broader values) and to reveal what is concealed by the
positive sciences. This would be rational enlightenment at its best, for we
would then be casting the light of reason as widely as possible.
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
46
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
epistemic link with them that does not involve mysticism. The apparent
insurmountability of this problem might persuade one to abandon pla-
tonism altogether in favor of some form of nominalism. The nominalist
will at least not have the problem of explaining how knowledge of abstract
entities or universals is possible because on this view there simply are no
abstract entities or universals. There are only concrete spatiotemporal
particulars, and it is argued that however one ends up construing these
there will be no great mystery about how we could come to know about
them. One could work quite naturally, for example, with a causal account
of knowledge.
The problems for nominalism lie elsewhere. Nominalism just does not
appear to do justice to existing parts of mathematics, and especially to set
theory. Mathematical statements do not appear, prima facie, to be about
concrete spatiotemporal particulars, even though they may in some cases
be applied to such entities. The language of mathematics does not itself
mention such objects. So one of the first problems for the nominalist
is to explain how and why mathematics is really so much different from
the way it appears to be, and from the way it is taken to be by practicing
mathematicians. For example, it is a fundamental assumption of different
mathematicians at different times and places that they are discussing the
same number (e.g., the number ) in their research. Mathematical practice
suggests that there is an identity through difference here, but how could
this be possible on a nominalist view?
We are also supposed to believe that we are systematically misled by
the language of mathematics but not, for some reason, by language that
refers to physical objects, or to concrete spatiotemporal particulars. Nom-
inalist enterprises are reductive, for they propose schemes for reducing
the language of mathematics to the language of concrete spatiotemporal
particulars. The reductive schemes proposed even for elementary num-
ber theory, however, have turned out either to employ notions that resist
nominalistic treatment or to be rather far-fetched. A major barrier to a
nominalistic treatment of mathematics lies in the fact that many mathe-
matical propositions are about infinite sets of objects, such as the set of
natural numbers, but seeing how such propositions could be reduced to
a language in which only spatiotemporal particulars are mentioned is dif-
ficult. The assumption that there is an infinite number of spatiotemporal
particulars goes beyond what is needed in physical theory and may in fact
be false. On the other hand, one might try to introduce modal notions
into the reductive scheme to provide for at least a potential infinity of
natural numbers. Here too there are problems. A coherent nominalist
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
§ 3 Intentionality
It takes no great insight to notice that our mathematical beliefs are always
about something. They are about certain objects, such as numbers, sets,
functions, or groups, or they are directed to states of affairs concerning
such objects. This ‘aboutness’ or ‘directedness’ of mathematical beliefs
Act(Content) −→ [object],
where we ‘bracket’ the object in the sense that we do not assume that
the object of an act always exists. Husserl is famous for suggesting that
we bracket the object, and that we then focus our attention on the act
and act-content, where we think of an act as directed toward a particular
object by way of its content.
The contents of acts can be determined by considering ‘that’-clauses
in attributions of beliefs and other cognitive states to persons. Consider,
for example, the following expressions:
11 See especially Edmund Husserl Ideas I, §§ 36–37, 84, 87–102, 128–135. Also, LI, Investi-
gations I and V.
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
12 For Husserl’s specific criticisms of nominalism see especially LI, Investigation II; §§ 2, 7–
8, 14, 26 speak to the issue of taking language about mathematical objects and universals
at face value. See also Ideas I, §§ 1–26.
13 See the LI, Investigations I and V; and Ideas I, §§ 87–91 and 128–133.
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
what we have are acts which . . . set up new objects, acts in which something appears
as actual and self-given, which was not given, and could not have been given, as
what it now appears to be, in these foundational acts alone. On the other hand, the
14 On ‘free variation’ see “Philosophy as Rigorous Science”; Experience and Judgment (EJ),
§§ 87–93; and Ideas I, §§ 67–72.
15 Husserl’s views on formalization and formalism are scattered throughout the Philosophy
of Arithmetic, LI, Ideas, FTL, the Crisis, and various unpublished writings and letters. For
my remarks here, see especially FTL, §§ 73–93, 101–107, and LI, Investigation I, § 20.
16 The themes of genetic analysis, founding and founded acts, and acts of reflection and
abstraction are found throughout Husserl’s work. See, for example, LI, Investigation
VI, §§ 40–58; FTL, §§ 70–71, 82–91; EJ; and the Crisis, Appendix VI, “The Origin of
Geometry.” See also R. Tieszen 1989.
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
new objects are based on the older ones, they are related to what appears in the basic acts.
(LI, Investigation VI, § 46)
§ 4 Mathematical Objects
If mathematical language is to be taken at face value, then on Husserl’s
theory of intentionality our mathematical beliefs are evidently about ob-
jects such as numbers, sets, and functions, and not objects of any other
type. This fact, taken together with what Husserl says about the ‘ideal’
nature of mathematical objects, suggests that we must confront the pla-
tonists’ problem of explaining how it is possible to know about ideal or
abstract objects. Husserl’s solution to this vexing problem is quite in-
teresting. Let us approach it by considering his view of mathematical
objects.
Perhaps the single most important thing to say about the concep-
tion of objects of cognition in Husserl’s transcendental phenomenology,
whether the objects be mathematical or physical, is that they are to be
understood in terms of the ‘invariants’ or ‘identities’ in our experience.
Many aspects of our experience are variable and in a constant state of flux,
and out of this flux, or against this background of variation, we find that
certain invariants emerge.17 Physical objects are identities that emerge for
us through various sensory experiences or observations. Facts about such
objects, obtained by empirical induction and expressed in empirical laws,
are to be understood in the same way. Now Husserl argues that we also
find invariants or regularities in our mathematical experience although
in mathematics the facts are not established, strictly speaking, by empiri-
cal induction. Mathematicians are not satisfied, for example, with estab-
lishing the truth of Goldbach’s conjecture or Fermat’s last theorem on
17 See, e.g., FTL, §§ 61–63; Cartesian Meditations (CM), §§ 17–18, 21; and EJ, §§ 13, 64–65.
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
objects, because social and cultural objects are bound to times and places.
Thus, Husserl says that mathematical objects are not ‘bound idealities’
but, rather, ‘free idealities’.20
Objects such as numbers are also identities that transcend conscious-
ness in the sense that there are indefinitely many things we do not know
about them at a given time, on the analogy with our knowledge of per-
ceptual objects, but at the same time we can extend our knowledge of
them by solving open problems, devising new methods, and so on. They
transcend consciousness in the same way that physical objects do. And,
similarly, we cannot will them to be anything we like; nor can we will
anything to be true of them. They are mind independent. On Husserl’s
view, unlike some empiricist or pragmatist views, it is not a puzzle that
experimental methodologies in the physical sciences are different from
methodologies in pure mathematics. Husserl was attempting to account
for the fact, for example, that theorems of mathematics are not expressed
probabilistically, and that we do not find in pure mathematics any state-
ments referring to the space-time properties of the objects under consid-
eration. Husserl’s view also explains differences in the dynamics of the
growth of the natural sciences and pure mathematics.
The abstractness of mathematical objects, on Husserl’s view, implies
and is implied by the fact that mathematical objects are unchanging,
omnitemporal, and acausal. As we just noted, there is also a sense in which
mathematical objects are mind independent and transcendent. Husserl’s
position is nonetheless different from classical or naive platonism in a
number of important respects. Husserl’s is a phenomenological (and a
transcendental) view of objects, not a classical metaphysical view. This
means that objects are to be understood as invariants in the phenomena
or in our actual experience in mathematics, and yet they are given as
objects that would exist even if there were no subjects. They are meant or
intended as objects that need not be grasped by any subject in order to
exist. The response to the nominalist is that the existence of mathematics
presupposes that there are identities through difference, except that these
are now simply understood as identities through the multiplicities of our
own cognitive acts and processes. They are not anything more ultimate
than that, not anything ‘metaphysical’ lying behind the phenomena. The
view is very much like what Kant has to say about empirical objects and
empirical realism, except that now it is also applied to mathematical
experience. On the object side of his analysis Husserl can still claim to be a
26 See Hermann Weyl Das Kontinuum 1918a, and “Comments on Hilbert’s Second Lecture
on the Foundations of Mathematics” 1928. Also, Oskar Becker 1927; Arend Heyting
“The Intuitionist Foundations of Mathematics” 1983; and Per Martin-Löf 1987. See
Chapter 13.
27 See R. Tieszen 1984.
28 See R. Tieszen 1994a.
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
29 On the concept of the ‘horizon’ of an act see Ideas I, §§ 27, 44, 47, 63, 69, 82–83, 142; EJ,
§ 8; CM, § 22.
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
§ 6 Against Fictionalism
Many basic objections to fictionalism about mathematical objects, and to
assimilating mathematics to fiction, fall out of what has already been said.
We have seen that Husserl does not wish to be understood as a ‘naive’
realist or platonist about mathematical objects, but he is also not a fic-
tionalist. Husserl says that “It is naturally not our intention to put the
being of what is ideal on a level with the being-thought-of which characterizes the
fictitious or the nonsensical” (LI, Investigation II, § 8). Let us consider some
of the objections that can be raised to fictionalism. Our intentions toward
some mathematical objects are fulfillable, but it is not clear what it could
mean for intentions to fictional objects to be fulfillable. Even where our
intentions to mathematical objects are not presently fulfillable we should
hesitate to assimilate these intentions to intentions expressed in fiction.
Generally, the idea that mathematical objects are fictions runs contrary
to the way that these objects are intended in the science of mathematics.
For example, various kinds of existence statements abound in mathemat-
ics texts, and it is not the business of the philosopher of mathematics
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
to explain these away. They are to be taken at face value, and it is the
business of the philosopher to explain how they are possible. In texts
concerning fictional objects we do not, however, find sets of statements
asserting the existence of objects. If the concept of existence even ap-
pears in such a context, it is subject to various anomalies that are not
present in mathematical language. A very substantial difference between
mathematics and fiction is suggested by Husserl’s discussion of reason
and actuality:30 a sustained belief in the existence of fictional objects, as
distinct from mere imagining of such objects, is a sign of irrationality. It
is a psychopathological phenomenon. On the other hand, belief in the
existence of mathematical objects, in the sense in which one believes ex-
istence theorems one has read in a mathematics text, is not at all a sign
of irrationality. On the contrary, mathematics is usually taken to be one
of the finest achievements of reason.
Other differences are also manifest. The idea of obtaining contradic-
tions from the assumption of the existence of certain objects, which shows
how such ‘objects’ do not have any stability in our mathematical expe-
rience, has no analog in our understanding of fiction. There is also a
disanalogy concerning questions of the development of knowledge and
of the determinacy of mathematical intentions when compared to inten-
tions in fiction. Thus, when we ask, “Was Hamlet over six feet tall or not?”
we find that no determinate answer can be given. Of course we can ex-
tend the story of Hamlet however we like so that we get a determinate
answer, but this is not at all how open problems are solved in the science
of mathematics. It could also be argued that mathematics ought not to
be assimilated to fiction because mathematics has extremely rich and
fruitful applications. It is not clear how this would be possible if math-
ematics were nothing but fiction. We might generalize from the point
Kant made about empirical objects: mathematical objects ought not to
be construed as ‘fictions’ just because we have no God’s-eye view of them,
or because it would be wrong to understand them as noumenal ‘things-in-
themselves’.
31 The concept of the ‘lifeworld’ in Husserl’s later philosophy is an outgrowth of his earlier
ideas on founding and founded acts, genetic analysis, and related themes. See especially
the Crisis, §§ 9, 33–38, Appendix VII.
P1: IWV
052183782c02.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 10:29
§ 8 Conclusion
The views described here are framed in Husserl’s later philosophy by a va-
riety of interconnected arguments against empiricism, naturalism, nom-
inalism, and psychologism in mathematics. It is clear from comments
in the “Prolegomena to Pure Logic” and other parts of the Logical In-
vestigations that Husserl sees nominalism and psychologism as species
of empiricism or naturalism about mathematics. They are forms of em-
piricistic reductionism, and Husserl is generally opposed to reductionism
in the philosophy of mathematics. In his criticism in Philosophy of Arith-
metic of Frege, who is certainly no empiricist about mathematics, we see
Husserl pointing out that Frege’s extensional definition of number also
amounts to a kind of reductionism about the meaning of the concept of
number.32 Husserl is concerned that any of the views mentioned not do
justice to our mathematical intentions, and to the way in which mathemat-
ical objects are intended in our acts. The views simply do not do justice
to what we know in the science of mathematics. For the same general
reasons, his view would be opposed to conventionalism, and to any kind
of strict formalism about mathematics. His views on formalism do not
entail that formalization in mathematics has no value. On the contrary,
formalization of our mathematical concepts is quite important. However,
from everything that Husserl says about intentionality, meaning, formal
(as opposed to transcendental) logic, the idealizations and abstractions
from experience required by formal systems, and the kind of informal
rigor which makes mathematics possible, it is clear that the later Husserl
is not a Hilbertian formalist.
Edmund Husserl is perhaps the only philosopher of the past one hun-
dred years or so who claims that we can intuit essences and, moreover,
that it is possible to formulate a method for intuiting essences. Husserl
calls this method ‘free variation in imagination’ or ‘ideation’. It is expli-
cated in some of his writings as the ‘eidetic reduction’. His descriptions
of ideation can be viewed as attempts to describe a method appropriate
to the a priori sciences, a method that does not reduce to the methods of
the empirical sciences. The best and clearest examples of this method, it
seems to me, are to be found in mathematics. Pure mathematics, accord-
ing to Husserl, is concerned with exact essences. Husserl’s own examples
of the method, however, are often concerned with the inexact (or ‘mor-
phological’) essences of everyday sensory objects (e.g., color, sound) or
of the phenomena that form the subject matter of phenomenology it-
self (consciousness, intentionality, and the like). It is unfortunate that
Husserl does not give more examples involving mathematics. He seems
to focus on the nonmathematical cases because he is very concerned to
show that phenomenology itself can be a kind of science, a descriptive
science of the essential structures of cognition. Perhaps he thinks it is
not necessary to dwell on mathematics, which, unlike phenomenology, is
already a firmly established science.
Husserl’s views on intuiting essences have been subjected over the
years to extensive discussion, criticism, and even ridicule. What I would
like to do in this chapter is bring a fresh perspective to bear on these
I would like to thank a referee for Philosophy and Phenomenological Research for comments on
this essay.
69
P1: IWV
052183782c03.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:5
… …
−4 −3 −2 −1 0 1 2 3 4
figure 1
does not remain invariant. Indeed, this can be made more precise in the
language of modern geometry: the origin and directions are invariant
under the group of transformations x = ax (where a is positive).
These simple examples show us that what remains invariant is a func-
tion of the kinds of transformations or variations we make. If the variations
are radical enough, we might expect that fewer invariant properties will
remain. Are there, for example, variations under which neither distance
nor direction would remain an invariant property? In fact, there would be
such variations, but then we can ask what would remain invariant under
these more radical variations. There will be, as it were, deeper or more
abstract invariants that we cannot eliminate even under the most extreme
imagined variations. We should therefore distinguish different levels of
invariants in relation to different groups of variations.
Husserl says in many works that the intuition of essences based on
free variation is not at all mysterious. Indeed, if the reader understands
the examples just presented, then he or she has just had such an expe-
rience. If you can grasp, that is, that distance is (not just hypothetically
but actually) invariant in our first example, then you have grasped an
essence. If you can see that direction is invariant in either example, then
you have intuited another essence. The invariants we have spoken of just
are essences in the sense that an ‘essence’ is a feature or property that
remains the same through many variations. It is something that a multi-
plicity of particulars have in common and is in this sense a universal. If
this is what we mean by ‘essence’ or ‘universal’, then we have in modern
geometry and topology a domain of cognition in which there already ex-
ists a highly developed mapping out of essences (or universals) and their
relations to one another (see § 2). A very sophisticated classification and
unification within geometry is indeed brought about in this manner, in-
cluding a sorting of geometric essences into species/genus hierarchies.
The whole of modern geometry is full of the kinds of examples we have
given.
We can thus see why Husserl speaks as though the experience of
essences is or could be quite commonplace. In laying out these examples
we have also not introduced any heavy metaphysical baggage into our dis-
cussion. Just imagine yourself engaging in the practice of geometry in this
manner (as would be common in modern geometry) without getting es-
pecially philosophical about what you are doing. Then we might say with
Husserl that you are apprehending an eidos or ‘idea’ (in something like
a platonic sense) but in its purity and free of metaphysical interpretation.
You are therefore apprehending it exactly as it is given immediately and
P1: IWV
052183782c03.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:5
body of work in mathematics that had its origins in the theory of alge-
braic invariants of Cayley and Sylvester, and even earlier work centering
on the concept of invariance, along with the subsequent developments
in geometry due to Grassmann, Riemann, Lie, Helmholtz, and others.
Husserl himself refers to these latter figures in his work on geometry.
There is, in my view, a sense in which we can see a preoccupation with
‘ideal’ invariants in geometry right from the beginnings of this science,
albeit in a much cruder form (see § 4 in this chapter).
What is nice about the approach to geometry developed by Klein, on
my view, is that it meshes quite naturally with Husserl’s descriptions of his
method of ideation once we consider it in the case of geometry. (Husserl
himself does not make this connection.) It provides a vivid illustration of
the method and does so in connection with an existing body of mathe-
matics. This does not mean that we should suppose that the method of
free variation in imagination must be understood everywhere in terms
of finding properties that remain invariant under groups of transforma-
tions. We need not necessarily use the group concept in all possible ap-
plications of ideation; nor need we import the specific technical aspects
in the geometric case back into the use of the method in other domains
(even in mathematics, to say nothing of ‘persons’, ‘consciousness’, and
the like). Perhaps we should view the idea of finding invariants under
groups of transformation in geometry as a specific instance of the method
of free variation in imagination. It stands to the general method in a
genus/species relation. It would be a variant of the method. We might
in some contexts appeal instead to what remains invariant under a ‘set’
or a ‘list’ of variations. The main idea is to work out all of the a priori
relations given whatever basic technical concepts we choose to employ at
the outset.
To elaborate somewhat on how structures of variation and invariance
are found throughout modern geometry, we can note that Euclidean ge-
ometry can be viewed as the study of those properties left invariant under
so-called rigid motions: translations, rotations, and reflections. Projective
geometry is concerned with the smaller class of essences that are a func-
tion of the rigid motions plus projections. Topology is concerned with
the still smaller class of essences that we obtain if our variations are even
more radical, including the most extreme stretchings and twistings. Thus,
length and angle are Euclidean essences, but they are not invariant under
projective variations. Linearity and triangularity are projective essences,
but they are not invariant under the more radical topological variations.
Connectedness and number of holes (to be more precise, ‘genus’), for
P1: IWV
052183782c03.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:5
example, are topological essences (see, e.g., Meserve 1953). One might
say that topological essences are quite abstract or deep, relative to other
geometric essences. They quite literally result from greater levels of ab-
straction. We can relate this idea of greater levels of abstraction directly
to the types of variations to which one subjects objects. In topology we
do not, for example, lose the property of dimensionality of a geometric
figure or object, or the property of having a boundary or not, but we
do lose properties such as size and shape as these are understood in Eu-
clidean geometry. Topological equivalence is much more abstract than
equivalence in, say, Euclidean geometry.
We can thus think of topology as a generalization of Euclidean ge-
ometry, and indeed of projective geometry. The group of topological
transformations has the group of projective transformations as a sub-
group and therefore has the group of Euclidean transformations as a
subgroup. The generalization/specification relations can in fact be made very
precise here. We can say exactly what is involved in making the geome-
try more ‘general’ or more ‘specific’ by pointing to groups of transfor-
mations (variations). Some variations are more radical than others. We
cannot perform them without changing a property that was invariant for
a range of variations into a property that is no longer invariant. The idea
is then to map out all of this, characterizing in each case the geometry
obtained.
There are properties that are essential to Euclidean space, but different
properties are essential to projective space, or affine space, hyperbolic
space, elliptical space, and so on. As Husserl says, an essence proves to
be that without which an object of a particular kind cannot be thought,
that is, without which the object cannot be imagined as such. In this
sense an essence is a kind of constraint that we cannot transgress. Thus,
Euclidean space and the objects in Euclidean space cannot be thought
without certain properties, projective space and the objects in projective
space cannot be thought without another set of properties, and so on. A
triangle or circle cannot be imagined to be just anything. Our imagination
here is not free but is bounded or limited in certain respects. If the
variations we make are too radical, we lose a property such as triangularity.
The essential properties will, as we said, be expressed in the axioms,
definitions, and theorems of each of the different geometries. We have,
for example, the hierarchy of geometries in modern geometry shown
in Figure 2 (viewing topology as a very abstract kind of geometry). An
essence of a given geometry is also an essence of every special case of the
geometry. A topological essence is also, for example, a Euclidean essence,
P1: IWV
052183782c03.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:5
topology
projective geometry
Euclidean
figure 2
projective geometry
affine
equireal similarity
Euclidean
figure 3
but there are many Euclidean essences that are not topological essences.
There are many more constraints associated with Euclidean geometry
than there are with the more ‘abstract’ geometries.
Note how this classification and organization extend to Euclidean and
non-Euclidean geometries. One can start with Euclidean geometry and
abstract until one obtains projective geometry or topological invariants,
or one can specialize from the top down to Euclidean geometry. In the lat-
ter case the number of invariant properties increases by specializing the
transformations under consideration until we have the invariant proper-
ties of Euclidean geometry. As we said, this is all made very precise in mod-
ern geometry by considering subgroups of a group of transformations.
To fill in the diagram in Figure 2 in even more detail, for example, we
might start with Euclidean geometry and remove the requirement that
area be invariant. Then we will necessarily lose some other properties as
well, and we obtain the group of similarity transformations. If we remove
the requirement that perpendicularity be invariant, we obtain equireal
transformations. If we remove both of these (abstract from them, as it
were) we obtain affine geometry. If we remove the restriction that paral-
lelism be invariant, we obtain projective transformations. These remarks
can be summarized as in Figure 3.
P1: IWV
052183782c03.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:5
n > 3. The reason is that in these cases we cannot make the objects the
theories are about present to ourselves in intuition in the same way that
we can with natural numbers and objects of ordinary Euclidean geometry.
We should thus note that intuiting essences and their relations to one
another, as we have been describing this, is very different from being
able to have the geometric objects the ontologies are about present to
us in sensory intuition. A cube, for example, is present to us in sen-
sory intuition in a manner quite different from that of a hypercube (a
four-dimensional object), and yet we can still develop the conception of
four-dimensional space and of what would have to be true in such a ge-
ometry. This is simply to acknowledge that our sense of perceived space
in dimensions beyond three falters quite dramatically. There is a shift
to analytic or algebraic methods and to questions about conceptual im-
possibility, possibility, and necessity. This is what Husserl is emphasizing
by highlighting the importance of the consistency of theories. It will be
worthwhile to say more about how this is all part of a phenomenological
account of the origins of geometry.
perfecting certain ‘limit shapes’ emerge toward which the series of per-
fectings tends, as toward invariant and never attainable poles. It is these
ideal shapes that make up the subject matter of pure Euclidean geometry.
Here we obtain an exactness that is denied to us in the intuitively given
surrounding lifeworld. It is the measuring of shapes in the prescientific
lifeworld that underlies these idealizations. The idealizations are a natural
outcome of refining and perfecting measurement. Every measurement
acquires the sense of an approximation to an unattainable but ideally
identical pole, that is, to one of the definite mathematical idealities or
to one of the numerical constructions belonging to them. It is in this
way that we obtain the ideality of basic geometric notions of points, lines,
planes, triangles, continuity, congruence, distance, direction, and so on.
Euclidean geometry is concerned with exact essences, and the awareness
of exact essences depends on idealization.
Euclidean geometry may thus be viewed as arising from the idealization
of structures given to us in everyday sensory perception. This geometry
existed for many years as a ‘material’ eidetic (a priori) science before it
was formalized. The arithmetization of Euclidean geometry constitutes a
very significant shift. Once numbers and the algebraic techniques of coor-
dinate geometry are brought to bear, the possibility of formalization arises.
We think of the plane, for example, as the set R2 of ordered pairs of real
numbers, R2 = {x, y : x, y ∈ R}, and a pair x, y represents a point in the
plane. A line is the graph of an equation of the form Ax + By = C where A
and B are constants that are not both 0. Slope, distance, and so on, are all
similarly characterized by formulae. With the subsequent formalization
of Euclidean geometry we obtain for the first time what Husserl calls the
‘Euclidean manifold’ of two or three dimensions. Husserl’s conception
of a manifold depends on a form/matter distinction. We abstract away
from the ‘matter’ or ‘content’ of geometric judgments to obtain the mere
form of the judgments.
Once we can represent points by pairs or triples of real numbers, lines
by algebraic formulas, and so on, it would be very natural to see whether
we could start thinking of a four-dimensional ‘space’ in which points are
given by quadruples of numbers w, x, y, z, the formulas for distance
and other properties are appropriately generalized, and so on. In other
words, the foundation is now in place for generalizing and for construct-
ing n-dimensional Euclidean or even non-Euclidean manifolds. One can
start using free imaginative variations on both formal and material as-
pects of the existing geometry (or geometries) to obtain new geometries.
In this manner Husserl wants to account for the origin and constitution
P1: IWV
052183782c03.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:5
in everyday sense experience begin to fail. But what if we could show that
the resulting geometries were indeed consistent? Then there would be
nothing illogical or incoherent about our conceptions in these cases. The
existence of relative consistency proofs of different geometries is just the
kind of thing Husserl would want to highlight in these cases.
Once we obtain new manifolds and formal systems through imagina-
tive variations and formalization, we can reinterpret these formal struc-
tures any way we like. We are free to reinstantiate them with ‘matter’ or
‘content’ in a variety of ways. Although Husserl does not discuss the topic,
one can say that this is just what happened in the case of the application of
non-Euclidean manifolds in relativistic physics, where we see a surprising
application of what had previously appeared to be merely conceptual or
symbolic mathematics (Weyl 1918b; Becker 1923; Mancosu and Ryckman
forthcoming). It is interesting to note, by the way, that at one time Einstein
evidently considered calling relativity theory ‘Invariententheorie’.
The Euclidean manifold and the Euclidean space for which it is the
pure categorial form are, for Husserl, prior to non-Euclidean manifolds
and spaces. So there is the already-characterized relationship between
the three-dimensional Euclidean manifold and the ‘space’ of everyday
sensory intuition. The three-dimensional Euclidean manifold can be ap-
plied in physics because it is a formalization of idealized geometry that
was arrived at on the foundation of everyday sensory experience in the
first place. Non-Euclidean manifolds might have applications to nature,
as happens in the case of relativity theory, but it does not follow that
our everyday sensory experience of the world is itself best characterized
in terms of such manifolds. Both kinds of manifolds might have appli-
cations to nature and we need to separate the question of applicabil-
ity from the question of characterizing of the space of everyday sense
experience.
Given this sketch of the origins of modern geometry we can start from
the bottom up or look from the top back down. In the one case we start
from the founding and more practical, concrete, sensory, and particular
and proceed to the founded and more theoretical, abstract, conceptual,
and universal; in the other case we proceed in the opposite direction. At
the deepest founding level one would consider the most basic, prescien-
tific space of everyday human practice, whereas at the highest founded
levels one would consider the most rarefied mathematical theories of
space. There are levels in the constitution of ‘space’ in which the sedi-
mentation of what comes before makes what comes after possible. There
are already invariants for us in sense perception – the objects themselves
P1: IWV
052183782c03.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:5
and some of their properties and relations – but then we see that there
are also invariants at ‘higher’ levels of cognition.
One might be tempted at this point to engage in what we might call
‘transcendental geometry’. With the very general post-Riemannian con-
ception of geometry in mind we might ask which of the indefinitely many
n-dimensional Euclidean and non-Euclidean manifolds are a priori con-
ditions for the possibility of our sense experience. Which are conditions
for our everyday experience of space? Could they all be? Only Euclidean
three-dimensional geometry? A related but different question is this:
which geometries are a priori conditions for the possibility of modern
physics?
We have seen in our characterization of the relationships of some of
the geometries that some geometric invariants are very deep and that
there are levels of invariants. Topological essences, for example, are
also Euclidean essences, but not all Euclidean essences are topological
essences. So one approach we might take, putting it very roughly, is just
to plug the hierarchical arrangement of invariants we have indicated in
our diagrams into our answer about the conditions for the possibility of
sense experience. We would then have a layering of conditions, in which
some are more abstract and general than others, and so on. The picture
that results is thus quite a bit more elaborate than the view found in Kant.
I do not, however, wish to go further into this subject in this chapter.
90
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
part ii
KURT G ÖDEL, PHENOMENOLOGY, AND THE
PHILOSOPHY OF MATHEMATICS
91
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
92
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
From the available evidence we know that Kurt Gödel began to study
Husserl’s phenomenology in 1959 (Wang 1978, 1981, 1987, p. 28). This
is an event of some significance for students of Gödel’s work, for years
later Gödel told Hao Wang that the three philosophers he found most
congenial to his own way of thinking were Plato, Leibniz, and Husserl
(Wang 1987, p. 74). Reports of Gödel’s interest in Husserl have also sur-
faced in other sources. Gian-Carlo Rota has written that Gödel believed
Husserl to be the greatest philosopher since Leibniz (Kac, Rota, and
Schwartz 1986, p. 177). And Heinz Pagels has written that “during his
later years he [Gödel] continued to pursue foundational questions and
his vision of philosophy as an exact science. He became engaged in the
philosophy of Edmund Husserl, an outlook that maintained that there
is a first philosophy that could be grasped by introspective intuition into
the transcendental structure of consciousness – the very ground of be-
ing” (Pagels 1988, p. 293). As part of his description, Pagels mentions how
Gödel thought it meaningful to question the truth of axioms, and to ask
about their philosophical foundations, and he then mentions Gödel’s
view on mathematical intuition. Georg Kreisel has also noted Gödel’s
I would like to thank Hao Wang for comments, and for discussion and correspondence
about Gödel’s philosophical interests. I have also benefited from correspondence with
Solomon Feferman, John Dawson, Jr., and Cheryl Dawson, and from comments by Lila
Luce, Penelope Maddy, Pieranna Garavaso, Steven G. Crowell, J. N. Mohanty, Izchak Miller,
and a referee for Philosophy of Science. Parts of this chapter were presented to the 1989 Eastern
Division meeting of the American Philosophical Association, to the 1989 spring meeting of
the Association for Symbolic Logic, and to the Philosophy Department Colloquium at the
University of Iowa. I thank members of those audiences for comments.
93
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
§1
According to Hao Wang, one of the logicians closest to the later Gödel,
Gödel remained interested in Husserl’s philosophy for a long time. Wang
has said that during his meetings with Gödel in 1971 and 1973 Gödel often
mentioned Husserl in conversation and urged him to study Husserl’s
post-1905 writings (Wang 1987, p. 120).1 Gödel was apparently interested
only in Husserl’s later, transcendental phenomenology, although he had
some reservations, to be discussed later, about Husserl’s last published
work, The Crisis of the European Sciences and Transcendental Phenomenology
(Crisis). Wang has also said that he and Gödel made a study of Husserl’s
“Philosophy as Rigorous Science.” A number of ideas about philosophy
as a rigorous science are included in Wang’s From Mathematics to Philosophy
(FMP) (1974, p. 6, pp. 352–356) and Reflections of Kurt Gödel (RKG) (1987,
Act(Content) −→ [object],
where we “bracket” the object because we do not assume that the object
of an act always exists. Phenomenologists are famous for suggesting that
we “bracket” the object, and that we then focus our attention on the
act (noesis) and act-content (or noema), where we think of an act as
directed toward a particular object by way of its content (or noema).
Husserl thought it possible to do a good deal of phenomenology on
the basis of simply reflecting on content without being concerned about
whether objects of acts existed or not. However, he also developed a
theory of knowledge on the basis of the notion of intentionality according
to which the question whether an object exists or not depends on whether
we have evidence for its existence, and such evidence is given in further
acts carried out through time. The phenomenological reduction has the
effect of making a belief in the existence of an object toward which an act
is directed dependent on the fulfillment, partial or otherwise, of (empty)
intentions directed toward the object, and one cannot impute existence
independently of this. An intuition is understood as a fulfillment of an
(empty) intention. It is the source of evidence.
Now from the “natural standpoint” one would assume that the object
toward which an act is directed exists. So the idea of suspending the nat-
ural standpoint amounts to not making this assumption, allowing that
there might not be an object, even though there is an intention directed
toward an object, and then understanding beliefs about whether an ob-
ject exists or not in terms of fulfillments of such intentions. This has many
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
2 On the basis of the preceding remarks about intentionality, this parallels an interpretation
of noemata due to Dagfinn Føllesdal.
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
cannot be made, are essential to the concept, and otherwise they are ac-
cidental. One might, for example, apply the method to the concept of a
mechanical procedure, or to the concept of number or set, or to various
geometrical concepts. Consider, for example, the concept of a triangle
and the properties that are or are not essential to this concept, given a
particular set of background assumptions. The method seems especially
appropriate for mathematical thinking, although with suitable distinc-
tions between material and formal essences, Husserl wished to apply it to
concepts of any type.
What remains invariant through the multiplicity of free variations is
supposed to be what is essential to a concept. The process of free varia-
tion is meant to foster the emergence of identities from or against a back-
ground of multiplicities, and “essences” are then understood in terms of
such identities or invariants. Husserl believed that through the process of
free variation one would uncover a “rule” governing the concept which
would not have been seen prior to the process of variation. The rule, even
if only partially understood or not fully determinate, would open up a
“horizon” for further possible insights.
Hence, the idea would be to attempt to understand (enough of) the
essence of the concept of set through the determination of new proper-
ties of sets to decide CH. In “WCCP?” Gödel had a particular concept of
set in mind: the “iterative” concept. Applying the method of free variation
to this concept could yield an axiom or axioms that would not be ad hoc
or artificial for the concept. Certain axioms would be “forced” upon us in
the sense that we could not imagine sets as objects in the cumulative hier-
archy without the properties expressed in the axiom(s). One might look
at the existing axioms of Zermelo-Fraenkel with Choice (ZFC) as axioms
that would be shown by free variation to be true of the iterative concep-
tion of set.3 This aspect of Husserl’s philosophy may be the foundation
for Gödel’s claim that just as we are forced in certain ways in perceptual
intuition, so certain axioms force themselves upon us as true of the con-
cept. One would not of course want a “decision” of CH to depend on an
3 One might try this as an exercise and a way to enter the frame of mind of using free
variation. There will no doubt be nuances and differences in evidence concerning some
of the axioms. In many of the expositions of the iterative concept, for example, it is
argued that replacement does not have the kind of immediate evidence that one finds
in the case of the other axioms. The axiom of choice is also set apart from the other
axioms in some treatments. Extensionality, unlike the other axioms, may be viewed as a
defining characteristic of sets (as opposed to properties). Much more could obviously be
said about this.
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
accidental property of sets, a property one could imagine sets’ not pos-
sessing. Hence, one might ask, for example, whether it is essential to the
iterative concept of set that sets be well founded, constructible, that there
be measurable cardinals, supercompact cardinals, that the axiom of de-
terminacy hold, or that the axiom of quasi-projective determinacy hold.
One can formulate axioms that decide CH, but at least at present none
of these is (intersubjectively) perceived as essential; they do not force
themselves upon us. How would we find new essential properties of sets?
Husserl’s method simply implies that further productive imagination is
needed. Productive imagination would be guided by the uncovering of
rules that would result from the process of free variation, given the ap-
propriate depth of understanding and background belief. But, in any
case, the idea that one can arrive at the essence of a phenomenon in this
way may appear to be a rather optimistic attitude toward open problems
that have been as vexing as the continuum problem. It appears to me
to be closely related to what Wang has described as Gödel’s “rationalistic
optimism.” Gödel’s rationalistic optimism appears to go hand in hand
with his realism or platonism, and with his view of the meaningfulness of
undecided mathematical statements.
Gödel was apparently also impressed by the phenomenological claim
that philosophy calls for a different method from science, and that it can
provide a deeper foundation for science by reflecting on everyday con-
cepts. Gödel was here presumably influenced by the role of what Husserl
later came to call the Lebenswelt in our scientific thinking. Nonetheless,
Gödel seems to have been very attracted to the idea that philosophy itself
would become a new kind of “rigorous science,” a science of essences.
Gödel apparently saw in Husserl’s idea of “philosophy as a rigorous sci-
ence” a revitalization of the Leibnizian ideal of philosophy. The tenor
of Gödel’s interest in Husserl in fact reflects the irony in how much of
Husserl’s philosophy was lost on his Continental followers, especially once
phenomenology took an “existentialist” turn at the hands of Husserl’s suc-
cessors. Husserl had of course been trained as a mathematician, and his
philosophy originally developed out of a concern for problems in the
philosophy of logic and mathematics. The sensibilities of a mathemati-
cian are apparent even in some of his very late writings. It is thus not very
surprising that Gödel did not care for parts of Husserl’s last published
work, the Crisis. It has been suggested by such philosophers as Merleau-
Ponty that the Crisis broke, at least tacitly, with the philosophy of essences
(I disagree with this assessment). In the language of the existentialists,
existence had begun to precede essence.
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
We can surmise from these comments that Gödel was primarily inter-
ested in the period of Husserl’s work in which Husserl believed that phi-
losophy could become a science. This is also the period in which Husserl
was a Kantian transcendental idealist who was nonetheless an “objectivist”
or a realist about mathematical objects. This last point is of some interest
because, as Kreisel has noted, Gödel was successful at mixing realistic and
idealistic conceptions in his own work in logic, mathematics, and physics
(Kreisel 1980, pp. 209–213). Gödel’s published and unpublished work
shows interests in both realism and idealism. In mathematics the inter-
ests in idealism are very clear in the papers on constructive mathematics,
whereas in physics one sees such interests in the papers on Kant and
relativity theory. Gödel had a long-standing interest in Kant even though
there were aspects of Kant’s philosophy that he did not like. Gödel (Gödel
1949a) argued that the models of Einstein’s field equations developed in
Gödel (1949) support an idealistic view of time and change in the sense
that these are to be viewed as a contribution of our own mind rather
than as an objective aspect of the physical world. Kreisel remarks that in
conversation Gödel did not view realism and idealism so much as conflict-
ing philosophies; rather, “he was ready to treat them more like different
branches of the subject, the former concentrating on the things con-
sidered, the latter on the processes of acquiring knowledge about these
objects or about the processes” (Kreisel 1980, p. 209). This attitude to-
ward realism and idealism is similar to Husserl’s attitude, and I will say
more about it later.
In the remaining part of this chapter I would like to pause over
Gödel’s comments on mathematical intuition in the 1963 supplement
to “WCCP?” for they have perhaps been his most widely publicized philo-
sophical views.
§2
Gödel’s most extensive published comments about mathematical intu-
ition are expressed in the supplement to “WCCP?.” Gödel speaks more
directly of intuiting sets than he does of intuiting or reflecting on the con-
cept of set, although the main body of the paper does contain references
to how reflections on the iterative concept of set “of a more profound
nature than mathematics is used to giving” (Gödel 1964, p. 257) are re-
quired to solve open problems in set theory. In the supplement, Gödel
is speaking about the intuition of transfinite sets, where one thinks of
such sets as objects in the (a) cumulative hierarchy. Gödel says that “the
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
In the next sentence of the passage Gödel says that the question of the
objective existence of the objects of mathematical intuition is nonetheless
an exact replica of the question of the objective existence of the outer
world.
In light of these kinds of comments, how is it that Husserl’s analysis
of intuition could be used to support or at least to help make sense of
Gödel’s views about mathematical intuition and the objective existence
of mathematical objects? Let us start with an example concerning ordi-
nary perceptual intuition. Perceptual acts, for Husserl, are paradigmatic
of intentional acts. Thus, they are acts directed toward objects by way of
their content, and the objects toward which they are directed need not
exist. The following example will help us to make a useful distinction
for a phenomenological understanding of mathematical intuition and
mathematical knowledge. Let us agree to speak of stages in our experi-
ence of objects. We should evidently think of the stages of our experience
as structured in linear time of type , although for my purposes at the
moment this will not be crucial. Now suppose I am staying at a cabin
in the mountains, look under my bed in the cabin, see a snake, scream,
“Snake!” and flee the room. Then at that stage of my experience I “see” a
snake. But suppose that what was really lying under the bed was a coiled
rope. Then, in a different sense of the word see, what I “see” at that stage
of my experience is a coiled rope. For the sake of continuity with Gödel’s
language let us use the terms see and intuit interchangeably. Then we
can distinguish these two senses of “seeing” or “intuiting” by adopting
some useful terminology that David Smith (Smith 1984) has introduced
in his study of phenomenology and theories of reference. Let us say that
I am representationally related to the coiled rope, but that, phenomenolog-
ically speaking, my experience has the representational character of being
“about” a snake. What I “intuit” at that stage of my experience is a snake.
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
future perceptions will yield. After some point the experience would pre-
sumably stabilize, just as our experience with sets will presumably stabilize
in various ways as research progresses. From this perspective it would also
be natural to speak of the objects of experience as “intentional objects.”
The intentional object in perception would be the object as it is presented
to us at a stage in our experience given the sensory data and the way in
which the sensory data are interpreted as a function of the intention,
background beliefs, memory, attention, and so on, at that stage.
We can read Gödel’s views on mathematical intuition as directly anal-
ogous to these views on perceptual intuition, with the qualification that
sensory data will not play the same role in mathematical intuition that
they play in perceptual intuition. Thus, Gödel does not say that we are
causally or representationally related to transfinite sets but rather that our
experience has the phenomenological character of being “about” such
objects. In fact, we could think of such objects as “intentional objects”
in the same way that we would think of ordinary perceptual objects as
intentional objects once we had taken a phenomenological approach to
perception. The question of the objective existence of objects of mathe-
matical intuition would then be an exact replica of the question of the
objective existence of the outer world: what would be relevant in both
cases would be whether we have evidence for the objects of our cognitive
acts as this would be provided in sequences of acts carried out through
time, whether we have fulfillment or verification procedures. As in the
case of the snake under the bed, our beliefs would be either verified or
not in further acts, and knowledge would accrue accordingly. This is why
Gödel could say in the 1963 supplement,
I don’t see any reason why we should have less confidence in this kind of per-
ception, i.e., mathematical intuition, than in sense perception, which induces
us to build up physical theories and to expect that future sense perceptions will
agree with them and, moreover, to believe that a question not decidable now has
meaning and may be decided in the future. (Gödel 1964, p. 268)
If Gödel was viewing matters this way, then he was, in effect, making
the phenomenological reduction. This conception of intuition would ex-
plain, as Gödel remarks in another passage, the analogy between illusions
of the senses and illusions in mathematics (Gödel 1964, p. 268). Gödel
remarks, for example, that “the set-theoretical paradoxes are hardly any
more troublesome for mathematics than deceptions of the senses are for
physics” (Gödel 1964, p. 268). The discovery of the set-theoretic para-
doxes showed that on the basis of the naive concept of set we were under
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
partial perceptions to see them that way. What is incomplete, partial, and
indeterminate at a particular stage is our knowledge or experience of the
object. The object itself is said to transcend our experience of it.
The phenomenology of perception thus shows that there is a robust
kind of “objectivity” associated with perceptual objects, and one could
say exactly the same thing about mathematical objects (or, if you like,
essences). In reasoning about numbers and sets we do not, for example,
suppose that these objects are somehow themselves indeterminate, in-
complete, or partial, although our knowledge about numbers and sets
certainly is in various respects indeterminate, incomplete, or partial.
Numbers and sets, moreover, are not thought of as objects that have tem-
poral characteristics, and they are not treated that way in mathematics,
but the acts in which we come to know about numbers and sets certainly
do have temporal characteristics. We can reason about the same number
in different acts, different people can reason about the same number, and
so on. Hence, the representational character of our experience presents
us with “an aspect of objective reality,” not something “purely subjective.”
This view also supports the platonist metaphor of knowledge as discov-
ery, for we do make discoveries about objects. We are firmly convinced in
many situations that “there is” another side to or aspect of an object even
though we do not presently see it. Something is there that we can come
to know about. We do not “create” what is there, so that in an important
sense it is not our own production. Similarly, in the case of open prob-
lems of mathematics we may be convinced that “there is” a solution to a
problem, or that there is an object of a certain type, even though we do
not yet know what it is. The process of free variation would enable us to
uncover a rule governing the concept(s) in question which would open
up a horizon of further possible insights.
The belief that problems have solutions, or that objects exist, even
though we do not yet know what they are, is connected with Gödel’s
“rationalistic optimism.” Wang reports, for example, that Gödel believed
Hilbert to be correct in rejecting the view that there are number-theoretic
questions undecidable for the human mind. For if the view were true, it
would mean that human reason is utterly irrational in asking questions
that it cannot answer while emphatically asserting that only reason can
answer them (Wang 1974, pp. 324–325).4 Human reason would then be
4 Gödel thought that one of the most important rigorously proved results about minds and
machines was the disjunction that either the human mind surpasses all machines, in the
sense that it can decide more number-theoretic questions than any machine, or there
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
very imperfect and even inconsistent in some sense, and that possibility
contradicts the fact that some parts of mathematics have been system-
atically and completely developed using laws and procedures that were
unexpected. Gödel cited the theory of first- and second-degree Diophan-
tine equations (the latter with two unknowns) and noted that solution of
all relevant problems in this area of mathematics supported rationalistic
optimism. Gödel’s rationalistic optimism goes hand in hand with his real-
ism or objectivism, and with his view of the meaningfulness of undecided
mathematical statements.
Our phenomenological observations also show how we could think
of idealism and realism as compatible in the way that Gödel evidently
thought they were. For unlike in naive forms of idealism or construc-
tivism, we are not claiming that we construct the object. That would be
a rather careless way of putting the matter. The correct way would be
to say that constructivism is concerned with the way the knowledge of
the object, not the object itself, is created or constructed. Idealism and
realism evidently are incompatible if idealism claims we construct the ob-
ject and realism claims we do not, or if idealism claims we construct our
knowledge of the object and realism claims we do not. I see this view of
the relationship between idealism and realism as reflecting just the kind
of twist on Kantian philosophy that one finds in Husserl’s transcendental
phenomenology: transcendental idealism coupled with mathematical (as
distinct from empirical) realism. On the analogy with the treatment of
empirical objects in Kant, numbers and sets are not mental entities; nor
are they fictions; nor are they some kind of radically mind-independent
things-in-themselves.
Gödel says that the presence in us of the abstract elements underlying
mathematics may be due to “another kind of relationship between our-
selves and reality” (Gödel 1964, p. 268). We might read this as meaning
that the character of our experience of sets may have a coherent causal
explanation but that it is of a different, probably more complex kind
from the causal explanation involved in my representational relation to
the coiled rope. The explanation would have to be different because
Gödel does not take sets, unlike coiled ropes, to be objects in the exter-
nal world. Thus we could not be causally related to sets in the way that we
exist number-theoretic questions undecidable for the human mind (Wang 1974, p. 234).
Gödel evidently believed that the second disjunct ought to be rejected. At the end of the
chapter I will briefly mention how Gödel connected the idea that the human mind might
surpass all machines to the notion of intuition.
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
§3
I believe that this reading helps to make sense of Gödel’s views about
mathematical intuition, even though epistemological difficulties about
how best to understand the notion of evidence in connection with the
idea of transfinite objects or processes remain. The notion of mathemat-
ical intuition is of course also connected with other themes in Gödel’s
work of which we have not taken note. For example, Gödel apparently
believed that something about the notion of intuition might suggest that
the mind is not mechanical, in the sense that it ought not to be viewed as
a finite, discrete, combinatorial, syntax manipulator. This theme emerges
clearly in Wang’s notes of his discussions with Gödel in FMP and RKG,
and in Gödel’s very interesting but unpublished “Is Mathematics Syn-
tax of Language?” (now published in two versions as Gödel *1953/59).
It is perhaps again related to the basic fact that in our experience we
intuit objects as complete, whole, and determinate, even though what
we actually see at a given stage in our perception is only one aspect or
part of an object. Gödel may have had in mind that it is possible that if
intuition were nothing more than combinatorial operations on discrete,
finite configurations of such “aspects” or “parts,” we would not have expe-
rience as we know it. But intuition is not like that, and hence it is possible
that mind is not mechanical. In a note published in FMP Gödel claims,
among other things, that Turing’s claim that mental procedures cannot
carry any further than mechanical procedures is inconclusive because it
depends on the supposition that a finite mind is capable of only a finite
number of distinguishable states (Wang 1974, pp. 325–326). He says that
“what Turing disregards completely is the fact that mind, in its use, is not
static, but constantly developing” (Wang 1974, p. 325). There is much
more to be said about this aspect of Gödel’s philosophical view, but we
cannot go into it now.
It seems to me that a phenomenological conception of reference and
intuition is worth exploring in its own right for the insights it might
provide into difficult problems about mathematical knowledge. Indeed,
I have discussed many of the issues in this chapter in detail, but limited
to finite sets, in Tieszen (Tieszen 1989). The notion of representational
character, and the role that it has in a theory of intentionality, is especially
interesting in the case of mathematics, for with it we are given some
leeway, of a type not furnished by other views of mathematical knowledge,
in dealing with the question of how knowledge of mathematical objects
is possible.
P1: IWV/KJS P2: IWV
052183782agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:23
Postscript (1992)
After this essay was written I learned from Solomon Feferman that a
manuscript in which Gödel discusses Husserl and phenomenology had
been recently transcribed and that it would appear in Volume III of
Gödel’s Collected Works. (It is now published as Gödel *1961/?) The
manuscript, written around 1961, shows that Gödel had by that time
assimilated many Husserlian ideas and that any later work could have
been influenced by his study of phenomenology. I also discovered that in
the present essay I was led to anticipate a number of the ideas that Gödel
explicitly discusses in Gödel *1961/?.
P1: IWV/KAC P2: IWV
052183782c05.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:26
The publication of Volumes I, II, and III of Kurt Gödel: Collected Works
(KG:CW ) (Feferman 1986, 1990, 1995) marks a major event in the his-
tory of logic and the foundations of mathematics. The material included
in the volumes not only presents us with a picture of the great scope,
depth, and significance of Gödel’s accomplishments, but will also open
up new avenues of thought and research for future generations of lo-
gicians, mathematicians, philosophers, computer scientists, and others
who will find Gödel’s ideas on various subjects to be of substantial inter-
est. There is an abundance of material to be considered in studying these
books, including work on relativistic cosmology, Kant and the philoso-
phy of time, an ontological proof for the existence of God, and Gödel’s
abortive but interesting efforts late in his career to settle the continuum
hypothesis. In this chapter I will focus only on indicating Gödel’s main
philosophical theses about mathematics and logic; even with this limita-
tion, it will only be possible to scratch the surface.
The bulk of Gödel’s technical work in logic and foundations was
completed between 1929 and 1943. The most important results are the
completeness theorem for first-order logic (1929), the incompleteness
theorems (1931), and the theorems on the relative consistency of the
axiom of choice (AC) and the continuum hypothesis (CH) (1938–40).
From 1943 on, Gödel devoted himself almost entirely to philosophy, but
he did not publish most of his work. Volume III of KG:CW contains
I thank Solomon Feferman and Charles Parsons for comments on an earlier draft of this
essay.
112
P1: IWV/KAC P2: IWV
052183782c05.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:26
Much of this early work shows a great deal of concern for construc-
tivity and for degrees of constructivity in mathematics. Indeed, Gödel
had already published some important results on intuitionistic logic and
arithmetic in 1933 (see 1933e, 1933f). He is, however, also critical of
intuitionistic notions (e.g., the intuitionistic notion of proof) as not be-
ing “constructive and evident to a higher degree” (see *1933o, *1938a,
*1941, 1958, and 1972). In this early period Gödel is especially inter-
ested in finding ways to establish the consistency of arithmetic and other
parts of mathematics on grounds that are constructive but that extend
beyond Hilbert’s finitism. (It is clear, especially in *1933o, that he al-
ready appreciated the difference between finitism and intuitionism.)
The “Lecture at Zilsel’s” (*1938a) gives a fascinating glimpse into this
effort and prefigures a number of important ideas and results. For ex-
ample, the idea of using higher-type functionals to obtain a consistency
proof for classical arithmetic is already discussed in the lecture. By 1941
Gödel (see *1941) had developed the so-called Dialectica-interpretation
of arithmetic. He did not publish this material until 1958 (see Gödel
1958), and he continued to add philosophical comments to an English
translation of it as late as 1972 (see Gödel 1972). Remarkably, the no-
counterexample interpretation of theorems of arithmetic also appears in
*1938a.
P1: IWV/KAC P2: IWV
052183782c05.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:26
by abstract concepts, in this context, are meant concepts which are essentially of
the second or higher level, i.e., which do not have as their content properties
or relations of concrete objects (such as combinations of symbols), but rather of
thought structures or thought contents (e.g., proofs, meaningful propositions, and so
on), where in the proofs of propositions about these mental objects insights are
needed which are not derived from a reflection upon the combinatorial (space-
time) properties of the symbols representing them, but rather from a reflection
upon the meanings involved. (Gödel 1972, pp. 272–273)
Gödel also argues in other works that reflection on the thought structures
or contents associated with mathematical symbols is needed in order to
find consistency proofs and to decide meaningful, well-defined mathe-
matical problems. He even explores ‘abstract’, nonformalistic concepts
of provability and definability in some of his writings.
In another paper from the thirties (*193?) Gödel says that Hilbert’s
belief in the decidability of every clearly posed mathematical question
is not shaken by the proof of the incompleteness theorems. The incom-
pleteness theorems show only that something was lost in translating the
concept of proof as “that which provides evidence” into a purely formal-
istic concept. Gödel concludes that it is not possible to formalize math-
ematical evidence even in the domain of number theory. The problem
of finding a mechanical procedure for deciding every proposition of a
class for certain classes of mathematical propositions, however, is abso-
lutely unsolvable. Another way to put this, according to Gödel, is to say
that it is not possible to mechanize mathematical reasoning completely.
The claim that the bounds of mechanism and formalism are not to be
identified with the bounds of human reason appears in many of Gödel’s
philosophical papers. Similarly, Gödel suggests in many passages that the
incompleteness theorems show that (the bounds on) what can be known
P1: IWV/KAC P2: IWV
052183782c05.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:26
mind, in its use, is not static, but constantly developing, i.e., that we understand abstract
terms more and more precisely as we go on using them, and that more and more
abstract terms enter the sphere of our understanding. (Gödel 1972a, p. 306)
Therefore, although at each stage the number and precision of the abstract terms
at our disposal may be finite, both (and therefore, also Turing’s number of distin-
guishable states of mind) may converge toward infinity in the course of the application
of the procedure. (Gödel 1972a, p. 306)
theories. He says that, thus far, demonstrability and definability have been
defined only relative to a given language, and “for each individual lan-
guage it is clear that the one thus obtained is not the one looked for.” With
the concept of Turing computability, however, “one has for the first time
succeeded in giving an absolute definition of an interesting epistemolog-
ical notion, i.e., one not depending on the formalism chosen.” Although
Turing computability is merely a special kind of demonstrability or defin-
ability, it is “by a kind of miracle” not necessary to distinguish orders. The
diagonal procedure does not lead outside the defined notion. It is this
fact that should encourage one to expect something similar in the case
of demonstrability and definability in spite of such negative results as the
incompleteness theorems and the Richard paradox. Concerning demon-
strability, Gödel is evidently thinking of the nonformal, abstract concept
of proof he mentions in other work, where ‘provable’ is understood in
the sense of ‘knowable to be true’. In the incompleteness theorems for
arithmetic, for example, it is not the case that one diagonalizes outside
what is knowable to be true in the natural numbers with the formation of
each new Gödel sentence. One does, however, diagonalize outside each
given formal system.
Gödel published his set-theoretic results between 1938 and 1940.
“Russell’s Mathematical Logic” (1944), which contains the first published
expression of Gödel’s mathematical and logical realism, appeared a few
years later. Gödel remarks on the realistic attitude that Russell adopted
in some of his work, and he develops some analogies between mathemat-
ics and natural science that are suggested by Russell. In particular, Gödel
notes how Russell compares axioms of mathematics and logic with laws of
nature, and mathematical evidence with sense perception. Axioms need
not be evident in themselves. Their justification could lie in the fact that
they make it possible for the ‘sense perceptions’ to be deduced. Gödel
thinks this view has been largely justified by subsequent developments,
and that in the future we will find it even more convincing. He writes that
it has turned out that (under the assumption that modern mathematics is con-
sistent) the solution of certain arithmetical problems requires the use of assump-
tions essentially transcending arithmetic, i.e., the domain of the kind of ele-
mentary indisputable evidence that may be most fittingly compared with sense
perception. (Gödel 1944, p. 121)
Classes and concepts may, however, also be conceived as real objects, namely
classes as “pluralities of things” or as structures consisting of a plurality of things
and concepts as the properties and relations of things existing independently of
our definitions and constructions. (Gödel 1944, p. 128)
mathematics in the subjective sense (i.e., viewed as the system of all demon-
strable mathematical propositions). If there exists a finite rule producing
all of the evident axioms of ‘subjective mathematics’, then, Gödel says,
the mind would be equivalent to a finite machine. It would, however,
be a finite machine unable to understand completely its own function-
ing. Furthermore, if the mind were equivalent to a finite machine, then
not only would objective mathematics be incompletable in the sense of
not being contained in any well-defined axiom system, but there would
exist absolutely unsolvable Diophantine problems. Absolutely here means
unsolvable by any mathematical proof the human mind could conceive,
and not just undecidable within some particular axiom system. According
to Gödel, the following disjunctive ‘theorem’ about the incompletability
of mathematics is inevitable:
Either mathematics is incompleteable in this sense, that its evident axioms can
never be comprised in a finite rule, that is to say, the human mind (even within the
realm of pure mathematics) infinitely surpasses the powers of any finite machine,
or else there exist absolutely unsolvable diophantine problems. (Gödel *1951,
p. 310)
of sensory intuition nor even with an eye to rounding out our theories
of sensory objects. Mathematics may be inexhaustible, but negative re-
sults such as the incompleteness theorems do not imply that there are
absolutely unsolvable mathematical problems. These results do not ex-
hibit a definite problem that cannot be solved. Rather, they establish the
nonexistence of a certain kind of general procedure for deciding a class
of problems. To a rationalistic optimist such as Gödel, this does not imply
absolute undecidability. It is empiricists, skeptics, and materialists who
are more likely to interpret it as an absolute undecidability result. Gödel
thinks it would imply absolute undecidability only if the mind were a finite
machine, but he gives arguments in various writings to show that the mind
is not a finite machine. We should keep in mind Gödel’s explorations of
nonformal, absolute notions of demonstrability and definability.
There is no doubt room to attack some of these ideas, but there is just
as much room to develop and defend many of them. In my view, Gödel
has a better philosophical perspective on mathematics than can be found
in much of the contemporary work on the subject. Of course there are
certain perplexing aspects of his views on mathematical intuition and re-
alism, but the perplexity seems to me to embody more insight than many
of the supposed (reductionistic) ‘solutions’ to these problems. I agree
with Thomas Nagel’s observation that there is a persistent temptation to
turn philosophy into something less difficult and more shallow than it
is. Part of the perplexity engendered by these views of Gödel probably
results from the fact that certain areas of philosophy have been neglected
in our time. For example, there has been almost no work at all on eluci-
dating a nonlogicist, nonreductionistic theory of concepts or properties
in connection with mathematics. As another example, one could point to
the fact that generations of analytic philosophers have ignored the work
of thinkers such as Husserl.
After studying these papers, it is easy to believe that Gödel’s ideas
will open up some new directions in the philosophy and foundations of
mathematics.
P1: IWV
052183782c06.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:36
I would like to thank the spring 1997 Logic Lunch group at Stanford, especially Solomon
Feferman, Grisha Mints, Johan van Benthem, Aldo Antonelli, and Ed Zalta, for help-
ful comments. I would also like to thank Andreas Blass, Albert Visser, and an anony-
mous referee for the Bulletin of Symbolic Logic for comments on an earlier draft of this
essay.
125
P1: IWV
052183782c06.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:36
to study any kind of objectivity whatever according to its general essence . . . means
to concern oneself with objectivity’s modes of giveness and to exhaust its essential
content in the process of “clarification” proper to it. . . . With this we meet a
science of whose extraordinary extent our contemporaries have as yet no
concept: . . . a phenomenology of consciousness as opposed to a natural science
about consciousness.
the spell of the naturalistic point of view . . . has blocked the road to a great science
unparalleled in its fecundity. . . . The spell of inborn naturalism also consists in
the fact that it makes it so difficult for all of us to see “essences”, or “ideas” – or
rather, since in fact we do, so to speak, constantly see them, for us to let them have
the peculiar value which is theirs instead of absurdly naturalizing them. Intuiting
essences conceals no more difficulties or “mystical” secrets than does perception.
(Husserl 1965, p. 110)
P1: IWV
052183782c06.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:36
But one must in no instance abandon one’s radical lack of prejudice, prematurely
identifying, so to speak, “things” with empirical “facts.” To do this is to stand like
a blind man before ideas, which are, after all, to such a great extent absolutely
given in immediate intuition. (Husserl 1965, p. 146)
According to Husserl, there are many ways in which essences have been
‘absurdly naturalized’. It is important to note that much of what has gone
by the name phenomenology since Husserl has itself naturalized essences
and shunned the ideal of philosophy as rigorous science. This devel-
opment began immediately with Husserl’s most famous and influential
student, Heidegger. It goes without saying that Husserl was not pleased
with this development.
Interestingly, Husserl laments many of the same reductionistic atti-
tudes about essences that are mentioned in Gödel’s discussions of math-
ematical content and abstract concepts: empiricism, naturalism, psychol-
ogism, nominalism, and forms of conventionalism and formalism that are
coupled with these views. Aristotelian realism may also be added to this
list. It is precisely this worry about reductionist attitudes that is a central
theme in Gödel’s 1961 paper. I now turn to Gödel’s remarks.
P. Bernays has pointed out on several occasions that, in view of the fact that the
consistency of a formal system cannot be proved by any deduction procedures
available in the system itself, it is necessary to go beyond the framework of finitary
mathematics in Hilbert’s sense in order to prove the consistency of classical math-
ematics or even of classical number theory. Since finitary mathematics is defined
as the mathematics of concrete intuition, this seems to imply that abstract concepts are
needed for the proof of consistency of number theory. . . . [What Hilbert means
by Anschauung is substantially Kant’s space-time intuition, confined, however, to
configurations of a finite number of discrete objects.] By abstract concepts, in
this context, are meant concepts which are essentially of the second or higher
level, i.e., which do not have as their content properties or relations of concrete
objects (such as combinations of symbols), but rather of thought structures or thought
P1: IWV
052183782c06.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:36
contents (e.g., proofs, meaningful propositions, and so on), where in the proofs of
propositions about these mental objects insights are needed which are not derived
from a reflection upon the combinatorial (space-time) properties of the symbols
representing them, but rather from a reflection upon the meanings involved.
The idea that it is necessary to reflect upon meaning plays a central role
in Gödel’s 1961/? paper, as does the idea that reflection on meaning
(or intuition of essence) is of a ‘higher level’ than reflection on the
combinatorial properties of concrete symbols.
There are proofs of CON(PA) and these proofs must therefore re-
quire objects or concepts of the sort that would be recognized by math-
ematical or phenomenological realists (see also Tieszen 1994b). That
is, the proofs must require abstract concepts and/or meanings that are
not available to concrete, sensible intuition. In addition, these objects
must in some sense be nonfinite. The requisite sense in which the ob-
jects or concepts needed for the proof of CON(PA) must be abstract
and nonfinite is seen in Gentzen’s consistency proof, since the proof re-
quires induction on the transfinite ordinals <ε0 . It is also seen in Gödel’s
consistency proof, since the theory of primitive recursive functionals re-
quires the abstract concept of a “computable function of type t.” If we
combine the conclusion drawn about meaning with the conclusion about
abstract elements, it appears that the meaning associated with arithmetic
expressions must be ‘abstract’, and that abstract elements cannot be elim-
inated from mathematics in the way that Hilbert had hoped. Another way
of putting this would be to say that Hilbert’s appeal to finite, concrete
particulars will not suffice for consistency proofs for interesting parts of
mathematics. In addition, the abstract elements involved could not be
given by Hilbert’s concrete intuition since concrete intuition is restricted
to finite sign-configurations. There must be, by the second incomplete-
ness theorem and the consistency proofs for PA, a less restricted kind
of mathematical intuition or insight that accounts for our mathemati-
cal knowledge. In Husserlian language, there must be an intuition of
mathematical essences. (This is no doubt what lies behind Gödel’s dis-
cussions with Wang (Wang 1974, pp. 84–86; 1987, pp. 188–192, 301–304)
about how we perceive or intuit abstract concepts, i.e., intuit categories
or essences.) It need not be claimed, however, that this kind of intuition
is infallible (see later discussion), even if it is taken to be a basic source
of evidence.
If the Gentzen or Gödel proof of CON(PA) is evident to us on the basis
of the meaning of the terms involved, then there is reason to believe that
P1: IWV
052183782c06.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:36
problem can be solved or they show that (ii) something was lost in trans-
lating the concept of proof as ‘that which provides evidence’ into a purely
formal or mechanistic concept. Gödel says it is easy to see that (ii) is true,
since number-theoretic questions that are not decidable in a given for-
malism are always decidable by evident inferences not expressible in the
formalism. The new inferences turn out to be exactly as evident as those
of the given formalism. Perhaps Gödel is overstating the case here. The
claim that number-theoretic questions not decidable in a given formalism
are always decidable by evident inferences not expressible in the formal-
ism seems to be true of questions such as that of the consistency of the
formalism (for reasonable formalisms), but it is not so obvious for arbi-
trary number-theoretic questions. Nonetheless, the conclusion we should
be able to draw is that Hilbert’s optimism about mathematical problem
solving remains untouched even though formalization or mechanization
of mathematical evidence in the domain of number theory is not pos-
sible. The reason Hilbert’s optimism remains untouched is that we can
hope to make these decisions on the basis of our directedness toward and
intuition of the underlying abstract concepts or essences. Gödel does not
say this in *193?, but by *1961/? it is clear that this is what he has in
mind (see § 6). Some parts of mathematics might be completely formal-
ized or mechanized, but, on the whole, it is not possible to mechanize
mathematical reasoning.
have been led by exploring aspects of this concept (see now also Hauser,
forthcoming).
Gödel goes on to say in *1961/? that it is not excluded by the incom-
pleteness results that every clearly posed mathematical yes-or-no question
is nevertheless solvable through cultivating our knowledge of abstract
concepts (or through developing our intuition of essences), for it is this
activity in which more and more new axioms become evident on the basis
of the meaning of the primitive concepts that a machine cannot emulate.
Gödel suggests in other writing (Gödel 1934, from the postscript added
in 1964; 1972a) that mental procedures may extend beyond mechanical
procedures because there may be finite, nonmechanical procedures that
make use of the meaning of terms. The intuition of mathematical essences
or concepts would be just such a procedure (see also Chapters 7 and 10).
Given the incompleteness theorems, there can for most mathematical
essences be no consistent machine that solves all of the well-defined yes-
or-no questions that are left undecided by the original sets of axioms for
those essences. Human reason, however, may be able to achieve such a
development by virtue of its ability to reflect on essences or concepts.
The mind can constantly develop without diagonalizing out of the math-
ematical essence it is intuiting. There might be a constant development
of machines to capture more of the essence but only by diagonalizing out
of each particular machine under consideration. Thus, it is through an
adjusted philosophical viewpoint according to which we intuit essences
that Gödel seeks to make a place for Hilbert’s optimism about mathemat-
ical problem solving as well as Hilbert’s idea that in mathematical proofs
we should strive for certainty.
Gödel says that the intuitive grasp of ever newer axioms that are log-
ically independent of earlier ones is necessary for the solvability of all
problems even within a very limited domain. He says that the appeal to
this kind of intuition is in principle compatible with the Kantian con-
ception of mathematics. Gödel points out, however, that Kant was wrong
to think that for the derivation of elementary geometrical theorems we
always need new intuitions, and that a logical derivation of these theo-
rems from a finite number of axioms is therefore impossible. In the case
of mathematics in a more general sense, however, Kant’s observation is
correct. Gödel says that many of Kant’s assertions are false if literally un-
derstood, but that they contain deeper truths in a more general sense. It
is Husserl’s (transcendental) phenomenology that for the first time does
justice to the core of Kantian thought. It avoids both the “death-defying
leap of idealism into a new metaphysics as well as the positivistic rejection
P1: IWV
052183782c06.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:36
of every metaphysics” (Gödel *1961/?). One could argue that the phe-
nomenological approach is not prey to the excesses and lack of balance
that characterized earlier ‘rightward’ viewpoints.
§ 7 Why Phenomenology?
Gödel worked mostly on philosophy after 1942. It is interesting to ask
why he would have settled on Husserl’s phenomenology and not some
other viewpoint. He certainly knew about extended forms of Hilbert’s pro-
gram, intuitionism, predicativism, and other foundational views. Quine’s
ideas had become very influential. He could have appealed to earlier
forms of platonism. There were even developments in post-Husserlian
phenomenology.
The answer to the question is straightforward. First, most of post-
Husserlian phenomenology had itself succumbed to leftward pressures.
Second, consider the views about provability, about what has meaning,
what objects can be recognized, and what can be intuited in (i) modified
forms of Hilbert’s program, (ii) traditional intuitionism, and (iii) pred-
icativism. To insist on restrictions such as those found in (i), (ii), or (iii) is
virtually to ensure there will be certain clearly posed mathematical prob-
lems that will not be solved, including questions about the consistency of
formal systems. It is, as Husserl might say, to ensure blindness or preju-
dice. In the case of the limitation to PRA, for example, it is to ensure that
the Gödel sentence for PA is undecidable, or that no consistency proof for
PA would be forthcoming. Or consider whether the following is a clearly
posed mathematical question: is Zermelo-Fraenkel (ZF) + continuum
hypothesis (CH) consistent or not? Adherence to the views (ii) and (iii)
would probably lead (or would have led) one to believe it is not a clearly
posed mathematical question, much less that it has a clear-cut solution.
On the basis of the meaning theories associated with these views, it would
be difficult to see how one could give meaning to the problem. In any
case, one would be blinded to the solution on views (i)–(iii) because the
proof that CH is consistent with the axioms of ZF requires impredicatively
specified sets. It also requires transcending of the intuitionistic ordinals
and acceptance of the classical ordinals as given. For some problems or
consistency proofs there is a need to ascend to higher types, new axioms
of infinity, classical ordinals, and so on, where this emerges naturally in
the course of unfolding the given essence.
Quine’s views on meaning, on what kinds of objects can be recognized,
and so forth, are beset with analogous problems (see Chapter 8). It will not
P1: IWV
052183782c06.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:36
help to be told that one can only recognize as legitimate the mathematics
that is needed to round out our theories of nature, much less that open
mathematical problems can or should be decided on the basis of whatever
mathematics this happens to be.
Most modern philosophical conceptions of mathematics are more
skewed in the leftward direction than ever before. Thus, the trend that
Gödel saw has not abated but has become stronger than ever. To such
philosophers as Husserl and Gödel this trend signals a crisis. Indeed,
it is one form of the crisis that Husserl writes about in The Crisis of
the European Sciences and Transcendental Phenomenology (see also Tieszen
1997b).
A few words should also be said about platonism. Husserl criticizes
earlier forms of platonism or realism, which he considers to be naive. A
central reason to avoid earlier, naive forms of platonism is that they place
essences outside all possible experience. They treat essences, in effect, as
abstract things-in-themselves. On the phenomenological view, however,
we are clearly directed toward and have access to essences. At least some
essences are here in our experience, if only partially. They are not all
completely outside or beyond our experience. To deny that essences are
here in our experience is simply to deny that we can grasp various cate-
gories in our experience without attempting to reduce them to something
else. This, in turn, is to deny the undeniable: that consciousness exhibits
intentionality.
I conclude by briefly considering a skeptical objection to the so-called
phenomenological method. The objection is this: what exactly is the
method supposed to be? How does one learn it? What are some examples
of the method? What are some of its fruits? My response is to argue that
these questions misplace the proper emphasis. For students of Gödel’s
work, the substantive point about the phenomenological approach lies in
its distinctive form of realism and its antireductionism about mathemati-
cal concepts and concept analysis. All questions about method should be
viewed in this light. Suppose, for example, that the arguments made
against psychologism by Frege and Husserl had not been successful.
Would thinking of logic and mathematics in terms of empirical psychol-
ogy have advanced logic and mathematics? Would moving logic and math-
ematics into the psychology department be an advance? It seems clear that
this would hardly be a desirable development. How would the ‘methods’
of logic and mathematics be conceived in those circumstances? One can
ask similar questions about strict formalism, mechanism, nominalism,
and other reductionistic and naturalistic viewpoints. In each case, the
P1: IWV
052183782c06.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:36
Parts of this essay were presented to the spring 2000 Working Group on the History
and Philosophy of Logic at UC-Berkeley and to the Spring 2000 Logic Lunch group at
Stanford. Thanks for comments are due to the audience members and especially to
Solomon Feferman, Grisha Mints, Paolo Mancosu, Tom Ryckman, Joel Friedman, Richard
Zach, Ed Zalta, Johan van Bentham, and John Etchemendy. I also thank the Synthese referees.
1 These arguments have been discussed by a number of writers. See especially Wang 1974,
1987, 1996; Parsons 1995a; Tragesser 1977; Tieszen 1994b; and Chapters 4, 5, 6, and 8 of
this book.
149
P1: IWV
052183782c07.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:56
Gödel says that there “exists today the beginnings of a science which
claims to possess a systematic method for such clarification of meaning,
and that is the phenomenology founded by Husserl.” Continuing, he says,
of some concept. Concepts, it seems, are objects sui generis. One could
make a list of things with which they are not to be identified or to which
they are not to be reduced. Gödel rejects those forms of reductionism that
ignore or seek to eliminate the intuition of concepts. He mentions specifi-
cally Hilbert’s formalism, mechanism, nominalism, conventionalism, pos-
itivism, empiricism, psychologism, and Aristotelian realism (Aristotelian
universals) (Gödel *1951, *1953/59, *1961/?). In § 9 of this chapter I
briefly consider some of the reasons for rejecting each of these positions.
Hao Wang recorded a remark from a conversation with Gödel that states
the point succinctly, albeit rather roughly (Wang 1996, p. 167):
Some reductionism is right: reduce to concepts and truths, but not to sense
perceptions. . . . Platonic ideas [what Husserl calls “essences” and Gödel calls “con-
cepts”] are what things are to be reduced to. Phenomenology makes them clear.
meaning or intention
A believes that S
subject act type is directed toward (refers to)
All higher forms of awareness involved in science and culture are in this
sense built up from reflection on everyday, prereflective, passive forms
of awareness. The existence of sciences such as mathematics and logic
indicates a more active and systematic cognitive achievement.
Concepts have the same basic function in these more theoretical
forms of awareness that they have in everyday awareness. Of course in
pure mathematics, unlike in everyday perception, the sensory constraints
on experience fall away. There are, however, still logical, conceptual,
or meaning-theoretic constraints on the experience. The concepts in-
volved simply direct us toward different kinds of objects in the two cases.
Following Husserl, we might also argue that the concepts of logic and
mathematics are exact in that they involve idealizations, whereas the
concepts involved in everyday perception are inexact or ‘morphologi-
cal’ since idealization is not involved in this case (see, e.g., Husserl Ideas
I, § 74). The points, lines, planes, circles, and spheres of Euclidean ge-
ometry, for example, are idealizations of the shapes of objects given to us
in everyday sense perception. It is possible to give a detailed account of
this cognitive activity of idealization, but there is no space to do so here
(see, e.g., Tieszen 2004).
Now certainly we can distinguish particular numbers from the concept
of number, particular sets from the concept of set, particular functions
from the concept of function, and so on. Imagine a cognitive life, for
example, in which particular natural numbers bore no more relation
to one another than they did to any other particulars. There could be
no systematicity for these objects. Each natural number would be utterly
unique and singular and could bear no more relation to another natural
number than to, say, a chair. Of course this idea flatly contradicts our
experience with natural numbers and it contradicts the fact that we have
a science of these objects. The concept of natural number is operative in
our experience with natural numbers. We also know, for example, from
our experience with natural numbers (as this is shown in mathematical
practice) that some concepts are not consistent with “x is a natural num-
ber” or that some are simply from the wrong categories to be applicable
to the objects that fall under this concept. An example of the latter type
would be a concept such as “x has a color.” Some of the concepts that are
consistent with “x is a natural number” are “x is even,” “x is prime,” and
“x is perfect,” Each of these concepts in turn has its own horizon. The fact
that the concepts have horizons and are not grasped with perfect clarity
is shown by the fact that there are open problems involving combinations
of these concepts. There are often various refinements and adjustments
over time in our intuition of concepts.
P1: IWV
052183782c07.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:56
The same points could be made about sets. However incomplete and
unfinished our thinking about sets may be, it nonetheless displays the
kind of systematicity and organization that indicate the presence of con-
cepts. Our early experience with the concept “x is a set,” for example,
appears quite indeterminate by later standards. The early experience has
been filled in with many concepts with various branchings, refinements,
and adjustments. Against this historical background we might single out
the concepts associated with Zermelo-Fraenkel set theory and expect new
concepts that consistently unfold the existing concepts to emerge. Such
new concepts would add further determinations that might allow us to
decide questions about these objects that were not decidable at earlier
stages. The situation here is analogous to the way that decisions may come
about through the progression of concepts involved in the perception of
the mandala. Gödel indeed says that the idea of deciding undecidable
mathematical propositions by extending mathematical theories with new
axioms or by ascending to higher types is best described as developing or
unfolding our intuition of the concepts (meanings) that appear in these
propositions (Gödel 1947, 1958, 1961/?, 1964, 1972). These decisions will
be nonarbitrary, as they are in the preceding example of sense percep-
tion. They will be forced or constrained in certain ways. Misperceptions
are also possible here.
As another example one might consider the concept of proof. The
intuitive concept of proof is clarified by the incompleteness theorems
themselves because they show that provability in any given formal system
cannot fully capture the intuitive concept (Wang 1974, p. 83). Purely
formal proof is always ‘relative’ to a given formal system. The incom-
pleteness theorems show us that number-theoretic provability is not the
same as number-theoretic truth, whereas at earlier points it was not clear
whether they could be equated or not.
Our examples show how concepts can become objects in reflection.
There is, in these cases, intuition of concepts. If we take intentionality
seriously, then we must say that in being directed toward a concept or category
of concrete particulars we are not (primarily) directed toward the particulars them-
selves. It is the concept that we grasp. The key idea to remember here is
that we are speaking of what the mind can be directed toward. Intuition is
to be understood only in these terms. It is a basic fact of consciousness that
we can be directed toward many different kinds of things, including, as
our examples are meant to show, concepts. If different categories of con-
cepts will be applicable depending upon whether we are directed toward
concrete particulars or concepts, then, according to the thesis of inten-
tionality, we must be directed toward different things. The properties
P1: IWV
052183782c07.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:56
and relations of concepts are indeed different from the properties and
relations of concrete particulars.
An immediate consequence of our remarks on directedness is that
directedness toward concepts is not the same as directedness toward sets.
The latter are governed by extensional identity conditions and the former
are not. We can of course also be directed toward sets in our thinking, as
is the case when we use the concept “x is a set” or some specification of
this concept (e.g., “x is a predicatively defined set”).
¬♦(∃x)(Px ∧ Qx).
Similarly,
and so on.
Consider again some of the previous examples. If the x in our ex-
perience is a dark round dot on a wall, then is it possible that x is a
mandala? Certainly these concepts are consistent with one another, but
the second concept is not implied by the first. On the other hand, the
concepts “x is a snake” and “x is a coiled rope” are not consistent with
one another. Note that this is not a purely formal contradiction, in the
sense of Px ∧ ¬ Px. One must know the origin of the symbols Px and
Qx. It is necessary to know what the symbols represent. To shift to some
mathematical concepts: if x is a natural number, is it possible that x is
P1: IWV
052183782c07.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:56
§ 5 Meaning Clarification
In the preceding discussion of intentionality it was said that we are
directed toward objects by way of the ‘meanings’ associated with our
cognitive acts. In reflection on experience we can explore and unfold
the meanings of our concepts. Thus, we are in effect describing a kind of
meaning clarification. One might seek to clarify, for example, the concept
P1: IWV
052183782c07.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:56
Qx?
yes no
yes no
Path 1 2 3
figure 5
P1: IWV
052183782c07.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:56
Frege liked to illustrate some of these ideas with his telescope analogy
(Frege 1892). Consider the observation of the Moon though a telescope.
The Moon is the object or reference (invariant), mediated by the real
image projected by the glass in the interior of the telescope, and by the
retinal image of the observer. The real image is like the sense (another
invariant), and the retinal image is like the subjective idea or experience.
The optical image in the telescope is one-sided and dependent upon
the standpoint of the observation, but it is still objective and invariant
inasmuch as it can be used by different observers or even the same ob-
server on different occasions. Each different observer, however, would
have his or her own retinal image as a result of the diverse shapes of the
observers’ eyes, different physiological features, and so forth.
Gödel appended to his 1972 paper. Gödel says that at one point Turing
presents an argument to show that mental procedures cannot go beyond
mechanical procedures. The problem is that there may be finite, nonme-
chanical procedures that make use of the meaning of terms. Turing does
not recognize that
mind, in its use, is not static, but constantly developing, i.e., that we understand abstract
terms more and more precisely as we go on using them, and that more and more
abstract terms enter the sphere of our understanding. (Gödel 1972a, p. 306)
Therefore, although at each stage the number and precision of the abstract terms
at our disposal may be finite, both (and therefore, also Turing’s number of distin-
guishable states of mind) may converge toward infinity in the course of the application
of the procedure. (Gödel 1972a, p. 306)
Human minds are not machines because human minds intuit abstract
concepts (see also Tieszen 1994b). Moreover, intuition of concepts in
my description is to be understood in terms of the intentionality of con-
sciousness, and intentionality is just what machines lack.
Nominalism is the view that only concrete particulars exist. Universals
do not exist, or, in the language I have used, abstract concepts/objects
do not exist. Nominalism is usually part of an empiricist epistemology
and ontology. Nominalists would certainly seek to eliminate the idea of
intuiting abstract concepts. According to the arguments in this paper,
however, we can say that concepts must exist in our experience in order
for it to be the way that it is. Either nominalism is a naive metaphysical
view that ignores the nature of our experience, or, if it does not ignore the
nature of our experience, then it is untenable. It is not the case that only
concrete particulars exist in our experience (see §§ 1–3). Gödel notes
that Hilbert’s formalism may be viewed as a kind of nominalism. In his
1953/59 papers he also shows how the incompleteness theorems can be
applied in particular to Carnap’s positivism, which can be viewed as a
combination of nominalism and conventionalism.
One of Gödel’s main objections (*1951, *1953/59) to conventionalism
is that it portrays mathematics as our own free creation or invention in a
way that does not square with the facts. As we said, not just anything falls
under a given concept. Not just anything could be, for example, a natural
number. We are constrained or forced in certain ways by the meanings
of concepts and we cannot change these meanings at will. There may be
a certain amount of freedom in unfolding a concept, depending on how
indeterminate the concept is. Conventionalism, however, misrepresents
this. It treats mathematics and logic as ‘human made’ and variable at will.
P1: IWV
052183782c07.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 11:56
intentionality and that the mind can be directed in many different ways.
The mind is not locked onto sense experience and only sense experience,
even if some elements of sense experience are always in the background
of consciousness. In particular, concepts make our experience possible,
and the mind can be directed toward these concepts.
Finally, Aristotelian realism is not necessarily a view that eliminates
intuition of abstract concepts. The problem is that it is just too limited
to do justice to mathematics and logic. It allows only those concepts or
universals that have physical, concrete instances. No other concepts or
universals exist. This is clearly not acceptable if we are to do justice to
pure mathematics and pure logic.
In light of more recent work on theories of concepts one could add
that concepts are not mereological sums, sets of actual or possible partic-
ulars, tropes, and so on. It might be possible to devise a formal theory of
concepts (see, e.g., Wang’s suggestions about this in Wang 1987, pp. 309–
311) but, as we have seen, Gödel makes various remarks about concepts
and meaning clarification that suggest the necessity to develop insights
into the content as well as the form of concepts.
About the process of the intuition Gödel says, in summary, that our
intuitions of concepts (1) are constrained or ‘forced’ in certain re-
spects (e.g., not just anything falls under the concept ‘natural number’);
(2) are fallible; (3) are more or less clear and distinct, precise; and (4)
are for many mathematical concepts inexhaustible (*1951, 1947, 1964,
and citations in Wang).
§ 10 Conclusion
Why would anyone think that there is or could be an intuition of abstract
concepts? I have now provided some of the elements of an answer to
this question. It is an answer that preserves some of the rationalistic,
‘rightward’ features (Gödel *1961/?) that have always been associated
with logic and mathematics while avoiding the metaphysical excesses of
naive rationalism and naive realism. Our experience in mathematics is
not random or arbitrary. It displays some degree of systematicity, which is
due to the concepts at work in experience. We should be able to uncover
additional relations among concepts. Viewed aright, the idea that we can
intuit concepts should hardly be the great mystery its detractors have
made it out to be. This of course is not to say that more work is not
needed.
P1: IWV/KAC P2: IWV
052183782c08.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:25
Charles Parsons (1995b, p. 309) has noted that Gödel never discussed
the deeper issues about meaning that are addressed by Quine. Parsons
says that the only place where Gödel even begins to approach these issues
is in the essay “The Modern Development of the Foundations of Mathe-
matics in the Light of Philosophy” (Gödel *1961/?). In it Gödel argues
that a foundational view that would allow us to cultivate and deepen our
knowledge of the abstract concepts that underlie formal or ‘mechanical’
systems of mathematics is needed. It should be a viewpoint that is favor-
able to the idea of clarifying and making precise our understanding of
these concepts and the relations that hold among them. Gödel says that
phenomenology offers such a method for clarification of the meaning
of basic mathematical concepts. The method does not consist in giving
explicit definitions but “in focusing more sharply on the concepts con-
cerned by directing our attention in a certain way, namely, onto our own
acts in the use of those concepts, onto our powers in carrying out our
acts, etc.” It is through such a methodological view that we might hope to
A version of this essay was presented at the Berkeley Logic Colloquium, April 1996, and
at the Stanford Philosophy Colloquium, May 1996. I thank members of both audiences
for comments, and especially Charles Chihara, Sol Feferman, Dagfinn Føllesdal, Thomas
Hofweber, David Stump, and Ed Zalta. I also thank Michael Resnik for comments and for
a spirited defense of some of Quine’s views. I doubt that he will be fully satisfied with my
responses.
In preparing this chapter, I have especially had in mind Charles Parsons’ writings on
Gödel and Quine, and some of his remarks on Husserl and Kant (Parsons 1980, 1983a,
1983b, 1990, 1995a, 1995b). Indeed, my essay can be read as a response to the comments
that Charles makes at the end of “Quine and Gödel on Analyticity” (1995b).
177
P1: IWV/KAC P2: IWV
052183782c08.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:25
Some reductionism is right: reduce to concepts and truths, but not to sense
perceptions. . . . Platonic ideas [what Husserl calls “essences” and Gödel calls “con-
cepts”] are what things are to be reduced to. Phenomenology makes them clear.
prejudices (see Husserl 1911, 1913). This surely explains Gödel’s remark
(Wang 1987, p. 193) that we might be able to see concepts more clearly if
we practiced the phenomenological epoché. This is an important part of
what it means to practice the epoché. It is not the silly, quasi-mystical un-
dertaking that some commentators have made it out to be. I will consider
later how Quine’s philosophy distorts mathematical content.
Total science is like a field of force whose boundary conditions are experi-
ence. A conflict with experience at the periphery occasions readjustments in
the interior of the field. Truth values have to be redistributed over some of
P1: IWV/KAC P2: IWV
052183782c08.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:25
Quine says that if his extended form of holism is correct, then it is folly
to seek a boundary between synthetic statements, which hold contin-
gently on sense experience, and analytic statements, which hold come
what may. Any statement can be held true come what may if we make dras-
tic enough adjustments elsewhere in the system. Conversely, no statement
is immune to revision, including statements of mathematics and logic.
The view that there is no interesting philosophical distinction between
mathematical and empirical truths reappears in many of Quine’s writ-
ings (see, e.g., Quine 1960, 1966a, 1970, 1974, 1992). There are only
gradations of abstraction and remove from the particularities of sense
experience in these truths but no sharp boundaries and no qualitative
differences or differences in type. Granted, mathematical content cannot
be understood along the lines of earlier, cruder forms of empiricism (e.g.,
Mill). It cannot, for example, be understood in terms of empirical induc-
tion. Instead, mathematical content will have to be more like the content
of the theoretical hypotheses of natural science. It will be more centrally
located in the holistic web of belief, less likely to be revised in the face
of recalcitrant experience, but in principle revisable. In this scheme, there
is no clear demarcation point of the the analytic, the a priori, or the
‘necessary’. Notions of mathematical necessity, possibility, and generality
are assimilated to natural necessity, possibility, and generality. Similarly,
there is no clear demarcation point of ‘certainty’, and the alleged cer-
tainty of mathematics will have to be understood accordingly. (I leave
aside discussion of the certainty of mathematics in this essay since I think
the issue is complicated by a number of factors.) In this context it will also
be difficult to understand the other rationalist element of mathematics
that Gödel mentions in 1961, which is the idea that a kind of ‘meaning
clarification’ based on conceptual intuition might play an important role
in facilitating the development of mathematics and in helping us to solve
open mathematical problems. In Quine’s work there is certainly no no-
tion of meaning clarification based on conceptual intuition. The very idea
of such a type of meaning clarification would be met with skepticism.
In contrasting Quine’s pragmatic holism with Gödel’s view I will argue
that it is the breadth of Quine’s holism that leads to problems. A holism
P1: IWV/KAC P2: IWV
052183782c08.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:25
§ 4 Analyticity
It will be useful to establish first a few simple reference points about the
notion of analyticity. We know that Gödel distinguishes a narrow from
a wide notion of analyticity and rejects the claim that mathematics is
analytic in the narrow sense. Gödel (1944, p. 139) says that ‘analyticity’
may have the purely formal sense that terms can be defined (either ex-
plicitly or by rules for eliminating them from sentences in which they
are contained) in such a way that axioms and theorems become spe-
cial cases of the law of identity and disprovable propositions become
negations of this law. In a second sense, he says, a proposition is called
‘analytic’ if it holds on account of the meaning of the terms occurring in
it, where meaning is perhaps undefinable (i.e., irreducible to anything
more fundamental). Gödel makes a similar distinction in a later paper
(Gödel *1951, p. 321), except that in this context he is thinking more
specifically of the notion of analyticity in positivism and conventional-
ism. In this later paper he says that to hold that mathematics is analytic
in a broad sense does not mean that mathematical propositions are “true
owing to our definitions.” Rather, it means that they are “true owing to
the nature of the concepts occurring therein.” This notion of analytic-
ity is so far from meaning ‘void of content’ that an analytic proposition
might possibly be undecidable. Gödel claims that we know about propo-
sitions that are analytic in the broad sense through rational intuition of
concepts. Analyticity is thus linked to a kind of concept description and
analysis.
Quine’s arguments in “Two Dogmas of Empiricism” and related works
are focused on a much narrower notion of analyticity than Gödel’s pre-
ferred notion. Quine looks to the views of Carnap and the logical posi-
tivists, and to the tradition that stems from accepting Hume’s distinction
P1: IWV/KAC P2: IWV
052183782c08.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:25
between matters of fact and relations of ideas. Gödel could perhaps even
agree with these arguments. Indeed, we should keep in mind Gödel’s
comments on what the incompleteness theorems show about the analyt-
icity of mathematics, namely, that mathematics could not be analytic in
the narrow senses that he indicates. If mathematics is analytic in the wide
sense, however, then relations between mathematical concepts must be
of a rather substantial nature. Determining relations between concepts,
as we described it earlier, must be different from determining purely for-
mal relations, relations of synonymy, explicit definition, convention, and
‘semantical rules’ of the type that Quine considers in “Two Dogmas.” At
the same time, it must not be based on sense experience and it is not
simply a function of rounding out our theories of sensory objects.
It is very important to keep in mind Gödel’s examples of proposi-
tions that are analytic in the wide sense. In particular, I think we must
start with examples in mathematics (and possibly logic) before we begin
to worry about whether there are wide analytic propositions containing
terms drawn from other domains of our experience. Gödel says in various
works that there exist unexplored series of axioms that are analytic in the
sense that they only explicate the content of the concepts they contain. An
example from foundations is provided by the same phenomenon Gödel
uses to refute Carnap’s idea that mathematics is syntax and is void of
content: the incompleteness theorems. On the basis of the incomplete-
ness theorems, an unlimited series of new arithmetic ‘axioms’, in the
form of Gödel sentences, could be added to the present axioms. These
‘axioms’ become evident again and again and do not follow by formal
logic alone from the previous axioms. Here we might say that by way of
a series of independent rational perceptions we are only explicating the
content of the concept of the natural numbers. The Gödel sentences we
obtain are compatible with this concept and do not overstep its bounds.
In the procedure for forming Gödel sentences, we do not diagonalize out
of this concept, even though we do step outside the given formal system.
Moreover, new propositions or axioms may help us solve problems that
are presently unsolvable or undecidable. One can already look at the un-
decidable Gödel sentence G for a formal system F in this way. We see that
F will prove neither G nor ¬G. But metamathematical reasoning shows
us that G is true if F is consistent, and we can thus ‘decide’ G on these
grounds. By adding this sentence to F we can create a new formal system
that will solve (albeit trivially in this case) a problem that was previously
unsolvable. This idea is related to Gödel’s comments on finding a view-
point that is conducive to solving meaningful mathematical problems
P1: IWV/KAC P2: IWV
052183782c08.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:25
does not come close to the goal of describing the cumulative hierarchy
with its membership structure since ZF1 has many nonstandard models.
For the intended interpretation we do better to look to second-order ZF
(ZF2 ) (compare, e.g., Kreisel 1967, 1971). (The possibility of viewing set
theory in this manner may be closed to Quine, given his strictures about
higher-order logic.) Although ZF2 is not itself categorical, its models are
known to be isomorphic to an inaccessible rank. Now on the basis of
the first incompleteness theorem one might suppose that the concept of
set-theoretic truth is richer than the concept of set-theoretic provability.
There will be truths of ZF2 , for example, that are not provable in ZF2 . In
particular, CH should have a truth value in the universe of ZF2 , even if
we do not presently know what it is. The truth or falsity of CH depends
on the breadth of the set-theoretic hierarchy and not on its height. The
relation of ℵ1 and 2ℵ0 is determined by the internal structure of the stages
of the hierarchy. One can therefore argue that the truth value of CH is
fixed by the contents of an initial segment of the hierarchy. By stage ,
sets of cardinality ℵ0 appear. By stage + 1, sets of cardinality 2ℵ0 ap-
pear. And by stage + 3, the pairing functions necessary for the truth of
CH will have appeared. In other words, there is some reason to believe
that CH should have a truth value under the intended interpretation of the
axioms of this theory. The intended interpretation is to be understood
in terms of the comments made earlier about intentionality. That is, it
is to be understood in terms of the idea that we are directed toward a
domain or universe by virtue of the meanings or contents of our acts, and
we can then further explore this domain. In this directedness we have an
example of what Gödel calls ‘rational intuition’. As we said, rational intu-
ition need not always be fully determinate. That is precisely why meaning
clarification is needed. We need not have and usually do not have a fully
determinate understanding of a domain from the outset. Moreover, there
are a variety of ways in which the intended interpretation might be cor-
rupted. We must be careful not to substitute some other content for the
given or intended content, even though this is what reductionist views
(such as empiricism and naturalism) would have us do.
a link that excludes intuition in the way that formal logic is supposed to
exclude intuition. Something outside the given logical formalism must
be involved. But we also do not learn about the link on the basis of sense
experience. This is why Gödel says we learn about it through ‘rational
intuition’. There must be a (partial) intuition of a concept (intention)
whereby the earlier axioms are related to the new axioms. There must be
a grasp of the common concept or concepts (intentions) under which
the axioms are unified. We can then begin to explore additional concepts
that consistently extend a given concept.
The notion of intuition of concepts is not mysterious if one is prepared
to recognize the fact that mathematical awareness exhibits intentionality.
As we said, the fact that there is directedness implies that there is catego-
rization in our experience. For an act to be directed in a particular way
means that it is not directed in other ways. Our beliefs are always about
certain types or categories of objects, and it is just these categories that we
are referring to as ‘concepts’. It is safe, for example, to say that we know
that certain things are not instances of the concept ‘natural number’ and
that other things are instances of this concept. Thus, we must have some
grasp of this concept even if our grasp is not fully precise and complete.
Instead of saying that we have a partial ‘grasp’ of a concept such as this we
might as well say that we ‘intuit’ the concept. The term intuition is used
because the concept is immediately given as a datum in our mathematical
experience once we adopt a reflective attitude toward this experience. It
is given prior to further analysis of the concept and to the consideration
of its relation to other concepts. The objections raised to Quine’s views in
this paper seem to me to point toward such a notion of rational intuition.
If it is possible to show that there are serious flaws in Quine’s view of
mathematics and if the exposure of these flaws seems to presuppose the
notion of rational intuition, then we have all the more reason to take the
notion of rational intuition seriously.
The view of analyticity just sketched is far from the tight little circle of
preserving or obtaining truths by synonym substitution or by the seman-
tical rules that Quine considers in “Two Dogmas.” The relation of one
axiom to another cannot be one of synonymy, or of ‘truth by semantical
rules’. Indeed, I shall suggest later that wide analytic truths share some
(but not all) features with the theoretical hypotheses of natural science
to which Quine generally wishes to assimilate mathematics.
Recognizing wide analytic truths is, in a sense, just a way of making
room for notions of meaning and the a priori that are needed to ac-
count for mathematical developments that have taken place since the
P1: IWV/KAC P2: IWV
052183782c08.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:25
for a formal system given that one is supposed to remain in the narrower
sphere of immediate concrete intuition of the type that is supposed to
accompany finitary mathematics (see Tieszen 1994b). The difference is
that in Quine’s case it is natural science, not finitist or constructive math-
ematics, that is supposed to provide the most secure and reliable basis
of knowledge. Natural science is epistemically privileged. The problem
is that it is privileged to the extent that it can blind us to the discovery of
consistency proofs, new axioms, and solutions to open problems, and to
the general development of mathematics.
To avoid these objections, Quine could try to revert to his set-theoretic
platonism. The problem with this maneuver, as we have seen, is that he
is then in a bind with his own views on natural necessity. He does in fact
make some effort to inactivate unapplied parts of set theory, as we will see
later. This effort suggests that he wants to backtrack on his platonism, but
without embracing constructivism. Another problem with the maneuver
is that Quine’s view of set theory does not embody a notion of content
that would allow us to avoid the basic objection.
The simplest known systematization today may not be the simplest known
systematization tomorrow.
It is also not clear how the idea of a simplificatory rounding off could
be compatible with advance to the more robust notions of mathematical
possibility and necessity that are required in the face of the incomplete-
ness theorems. Advancing to higher types or sets is a natural and intrinsic
extension of Gödel’s concept of set, but whether it answers to Quine’s
concerns for simplicity and economy with respect to natural science is
not clear. In short, it is not clear what the appeal to simplicity and econ-
omy could amount to in the presence of the incompleteness phenomena
and, more generally, of the inexhaustibility of mathematics. The kind of as-
cent needed to solve open mathematical problems takes us further from
natural science and its problems and has nothing to do with rounding out
the mathematics needed for the natural sciences. Empiricist accounts of
mathematics tie mathematical concepts more closely to acts of sense per-
ception in one way or another and thus place constraints on mathematical
thinking instead of fostering its expansion. On the other hand, acts in
which there is a free imagination of possibilities, when placed in the service
of reason, lead to a far less constrained view of mathematical possibility.
One has to ask how to proceed with research about open mathematical
problems. For Quine, the ‘simplificatory rounding out’ is always a round-
ing out of what is needed for natural science. What cannot be handled in
this way is treated as on a par with uninterpreted systems or mathematical
recreation and, as we see, is regarded as gratuitous. But what is gratuitous
or simplest with respect to the objective of solving mathematical problems?
One could argue that with such an objective in mind, deciding in favor
of V = L would be gratuitous and perhaps even contrary to canons of
simplicity.
In a sense, the problem goes beyond Quine’s view of higher set theory.
The problem is that Quine has no place at all for the intended meanings
of mathematical theories. Quine has on occasion said that a serious di-
vergence over logic or set theory is just a confrontation of rival formal
stipulations or postulates. He has said that we make deliberate choices
and set them forth unaccompanied by any attempt at justification other
than appeals to elegance or convenience. Although this may be true of
Quine’s attitude toward mathematics and set theory, it does not gibe with
the usual attitude toward these subjects. It is not accurate as a report on
mathematical practice.
It is worthwhile to consider the meaning and motivation of Quine’s
set theories ML and NF in light of these comments on rival formal
P1: IWV/KAC P2: IWV
052183782c08.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:25
§ 11 Conclusion
If the arguments in this essay are correct, then Quine fails to reconcile
mathematics with his pragmatic brand of empiricism. The attempt to ex-
tend holism across the fields of empirical science, mathematics, and logic
P1: IWV/KAC P2: IWV
052183782c08.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:25
I thank Michael Resnik and Charles Chihara for comments on this essay.
201
P1: IBE/JTR/NDO P2: IWV
052183782c09.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:28
objection since the analogy should break down if sets are not located in
space and time. Thus, I will consider the objections together.
First, concerning the ontological point, it is probably safe to say that no
argument is needed for the claim that a deck of cards is physical. At the
same time, we do not have a science made up of theorems about decks
of cards. One of the most striking characteristics of mathematics is that
its axioms and theorems, once established, seem to have a remarkable
stability over time. That is, investigations of the sort that lead to axioms
and theorems are taken by mathematicians to be about objects that are
identical across times, places, and persons. Different mathematicians at
different times and places believe they are dealing with the same object
(e.g., the set of prime numbers or the number ) in their investigations,
and this accounts for why the science of mathematics is possible at all.
And doing justice to mathematical practice requires taking such basic
beliefs of mathematicians into account. If mathematical objects were not
(believed to be) characterized by such invariance, mathematics would
be constantly shifting, with different objects and different theorems for
different times, persons, and places. We would see a kind of relativism
in mathematics that is alien to the way that this science is actually given.
(Naturalism, it is worth noting, has always been accompanied by some
form of relativism.) Now Maddy’s perceived ‘sets’ simply do not have this
kind of stability or invariance. They could not be such invariants in math-
ematical experience because, as noted previously, they change in many
ways. The origin of our awareness of sets may have something to do with
seeing groups of physical objects, but one could argue that only when one
has grasped this kind of invariant does one have a set as it is understood in
mathematics. So what is needed is an account of the knowledge of these
invariants. And Maddy’s views on generalization, presented later in the
book (Chapter 4), do not help (see later discussion). If these comments
are correct, then set theory could not itself be about physical things that
shift locations, blink in and out of existence, and are constantly changing.
It may have applications to these things, but even its applications involve
a certain degree of idealization.
This point is directly related to the objection that challenges the anal-
ogy between the perception of physical objects and the perception of sets,
for if sets are not objects in the environment to be detected by neural
cell-assemblies, one would expect the analogy to break down. Maddy’s
initial response to this objection is to point out that the gold-dubber has
in fact causally interacted only with the front surface of a time slice (an
aspect) of the sample and that, by analogy, the set-dubber also interacts
P1: IBE/JTR/NDO P2: IWV
052183782c09.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:28
of these choices is the correct one. Therefore, numbers are not sets. In
fact, numbers are not objects at all, because any objects we identify them
with will have additional properties which are superfluous to the object’s
numerical functioning. Again, there are no arguments that one of these
choices is the correct one. Therefore, numbers are not objects at all. But
then what is the nature of the ontological relationship between numbers
and sets?
Maddy accepts the conclusions that numbers are not sets and are not
objects. To continue her effort to naturalize mathematics, however, she
wishes to locate numbers in space-time. She holds that numbers are prop-
erties of sets, analogous to physical properties, in particular, to physical
quantities. Just as the perception of physical objects includes the percep-
tion of their properties, so the perception of sets does. She argues that
knowledge of numbers is knowledge of sets because numbers are proper-
ties of sets. Moreover, knowledge of sets presupposes knowledge of num-
ber since, for example, Piaget’s work indicates that subset relations can-
not be properly perceived before number properties. Arithmetic, from
this perspective, is part of the theory of hereditarily finite sets. The prob-
lem of multiple reductions of numbers to sets in set theory is now to be
viewed in the following way: consider the von Neumann ordinals. They
are nothing more than a measuring rod against which sets are compared
for numerical size. So the choice between the von Neumann ordinals and
the Zermelo ordinals, for example, is no more than the choice between
two different rulers that both measure in meters, and asking whether a
number is one of these kinds of sets is like asking whether an inch is
wooden or metal.
Maddy discusses objections to the identification of natural numbers
with properties of sets raised by Frege and Benacerraf and argues that the
effort to construct a neo-Benacerrafian dilemma for property theories –
which properties are really the numbers? – fails. Her argument, which I
do not find convincing, rests on the view that numbers are to be under-
stood in a particular way as scientific properties, not as sets or predicates.
She leaves open the question whether properties (as universals) should
be included in the set-theoretic realist’s ontology, pointing out that the
question of the ontology of properties is just as pressing for physical sci-
ence as it is for mathematics. Further discussion of a Fregean proposal for
what numbers are leads to some interesting comments on proper classes
and the nature of the distinction between sets and classes.
At the beginning of Chapter 4, “Axioms,” Maddy notes that her view
of the perception of sets gives us only the barest beginning of an account
P1: IBE/JTR/NDO P2: IWV
052183782c09.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:28
not consider the arguments that seem to force such a conclusion. In the
same vein, she is silent on the matter of the intentionality of perception
and belief and on the difficult problems, discussed by Quine, Putnam,
and others, that stand in the way of naturalizing intentionality.
Maddy holds that whereas axioms like Pairing and Union are sup-
ported intuitively, the other axioms of set theory must be treated differ-
ently. She focuses on the history of the axiom of Choice in particular and
analyzes it in some detail, arguing that it is supported in part by intuitive
evidence and in part by extrinsic means. It is not clear, however, how
the account of the intuitive support for Choice described in this chapter
jibes with the details of her own view of intuition discussed earlier, for
here she simply cites Zermelo’s remarks on the way the axiom was used
by mathematicians before it was explicitly formulated. She takes the ax-
iom of Infinity (‘there is an infinite set’) to be a bold and revolutionary
hypothesis. There is nothing intuitive about it, but it launched modern
mathematics, and the success and fruitfulness of that endeavor provides
its purely theoretical justification. (But was there not some sort of axiom
of infinity in ancient Greek mathematics, for example, in the idea that a
line is infinitely divisible?)
Maddy finds that a variety of extrinsic supports are in fact offered
by set theorists: appeals to ‘verifiable’ consequences, to powerful new
methods for solving preexisting open problems, to simplification and
systematization of theory, to implication of previous conjectures, to the
implication of ‘natural’ results, to strong intertheoretic connections, and
to provision of new insight into old theorems. She also discusses how
open problems in set theory, such as the continuum hypothesis, should
be understood from this perspective and finds, interestingly, that in the
second tier there are some arguments for new axioms that do not depend
on consequences, but also cannot be based on intuition. The ideas they
rest on, which she calls ‘rules of thumb’, cannot be easily classified as
either intuitive or extrinsic.
The line of argument involving the status of the empty set, unit sets,
and all those sets whose members cannot be construed as physical objects
is picked up in Chapter 5, “Monism and Beyond.” In this final chapter
Maddy responds to a number of objections that have been raised to her
views by Charles Chihara and tries to show that set-theoretic realism is
consistent with a physicalistic ontology. Here she says that anyone unaf-
flicted by physicalistic scruples is free to hold that we gain knowledge of
pure sets by theoretical inference from our elementary perceptual and
intuitive knowledge of impure sets. Why or how this should be so she does
P1: IBE/JTR/NDO P2: IWV
052183782c09.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:28
not say. In any case, physicalists will complain that this does not solve the
problem of spatiotemporal location, and so Maddy argues that pure sets
are not really needed. She says that the set-theoretic realist who wishes to
embrace physicalism can locate all the sets she needs in space and time.
All pure sets can be eliminated. Moreover, she argues that physical objects
can be identified with their singletons, so problems about unit sets evap-
orate. Maddy says that not every set-theoretic realist in her sense must be
a physicalist and that nonphysicalists may prefer to retain the ‘standard’
version of set-theoretic realism with its pure sets. But the identification
of physical objects with their singletons and the elimination of pure sets
yield a version of metaphysical monism in place of the customary dual-
ism of the mathematical and the physical. On this monistic view, every
physical thing is already mathematical and every mathematical thing is
based in the physical. This view is offered as an option to those for whom
it holds some appeal.
In the final sections of Chapter 5 Maddy compares the monistic (not
the standard) position to Hartry Field’s nominalism and finds monism to
be, on the whole, a better position. Both forms of set-theoretic realism are
also compared to and contrasted with the structuralist views of Michael
Resnik and Stewart Shapiro.
So we see in this final chapter that in order to solve the problems
at hand there is a splitting of set-theoretic realism. The split, however,
has the effect of straining if not unraveling the compromise platonism
that Maddy wants, for now there appear to be dualistic and monistic
forms of compromise Platonism. The monistic version is supposed to
solve problems about pure sets and unit sets, but in so doing it is subject
to some of the same objections that compromise platonism was meant to
avoid. For example, it is not clear that eliminating pure sets, or requiring
that physicial objects be identified with their singletons, is consistent with
taking set theory at face value, as compromise platonism was supposed
to do. If we do take set theory at face value, then there seem to be some
very simple and obvious truths into which only pure sets figure. How is
the obviousness of these simple truths to be explained by the monistic
version? This is a variant of the obviousness problem raised earlier for
the Quinean view. Moreover, it is not clear how the monistic version
recognizes a uniquely mathematical form of evidence for (unapplied
parts of) set theory.
The dualistic version, on Maddy’s own terms, is apparently not consis-
tent with physicalism. Otherwise, why offer the alternative? Perhaps this
is not a serious problem, but then how the dualistic version can handle
P1: IBE/JTR/NDO P2: IWV
052183782c09.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:28
problems about pure sets and unit sets is still not clear. Maddy says very
little about how we could gain knowledge of pure sets by theoretical infer-
ence from our elementary perceptual and intuitive knowledge of impure
sets. And it again appears that there will be an ‘obviousness’ problem for
simple truths involving pure sets since Maddy will now have to sort these
truths into the theoretical (versus intuitive) side of her epistemological
scheme.
Another problem for the compromise results from the two-tiered epis-
temological structure. How the two tiers are related is not clear. The first
tier is supposed to have perception of sets. The second tier does not have
perception of sets but, rather, extrinsic justification of the Quinean type.
The second tier is supposed to handle all of the sets that are routinely
taken to exist in set theory but are not perceivable (that is, the theoret-
ical part of mathematics). (This suggests, incidentally, that the meaning
of the existential quantifier in mathematics is not univocal but must be
understood differently, depending on whether it occurs in a statement
that is intuitively supported or extrinsically supported.) Now it appears
that even if we allow that there is sensory perception of sets, this per-
ception could not be a necessary condition for mathematical knowledge
in the second tier. To see this consider, for example, Maddy’s account of
the difference between the axioms of Pairing and Infinity. The analysis
of the epistemological basis of these axioms mentioned earlier suggests
that the two tiers are in fact logically independent of one another. It
could be argued that there is really a very radical division here. That is
why Gödel focuses on intuitive, intrinsically mathematical evidence in-
volving abstract objects even at higher levels of set theory and introduces
extrinsic criteria as an alternative. On the other hand, if you recognize
extrinsic criteria, then, as in Quine’s view, why not apply them uniformly
to all of the axioms? What could mathematical intuition possibly add? In
short, it is not clear that ‘compromise’ platonism is a coherent position
in the philosophy of mathematics.
Apart from these issues, Realism in Mathematics prompts a number of
general questions about why and even whether we would want mathemat-
ics to be naturalized. The claim that there is no point of view prior to or
superior to that of natural science is no doubt persuasive as part of a gen-
eral cultural perspective, that is, as a result of the great success of natural
science and of the technologies based on it. Nonetheless, there are some
well-known problems with the claim. Consider, for example, the follow-
ing questions: how we are to conceive of science? and how we are to view
the relationships between the various sciences (e.g., neuroscience and
P1: IBE/JTR/NDO P2: IWV
052183782c09.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 12:28
psychology)? It is not clear that these are questions that could or should
be answered by natural science. To attempt to naturalize mathematics is
perhaps inevitably to be a kind of reductionist about mathematics, and
this can have a number of negative consequences. It could actually hin-
der mathematical progress at the expense of some philosophical theory. I
think that Frege and Husserl had this concern in mind when they criti-
cized the prevalent form of naturalism of their time, psychologism, and
that Gödel also had it clearly in mind when he criticized various forms of
empiricism and positivism about mathematics. (See also his remarks on
why the completeness theorem for first-order logic was not discovered
earlier.) Focusing too heavily on inductive criteria or extrinsic justifica-
tion for the truth of mathematical statements, for example, might blind
us to interesting or important results that could be established intrinsi-
cally (e.g., on the basis of mathematical intuition, as in Gödel’s view). Or
trying to naturalize mathematics could lead to an exclusionist attitude
about forms of reasoning in mathematics that appear to resist such treat-
ment. There may be ways to naturalize mathematics, however, that are
not subject to these problems.
P1: ICD
052183782c10.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:14
10
215
P1: ICD
052183782c10.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:14
The central ideas in each part of the book are quite interesting, but
they are also highly controversial. I will focus on the material on Gödel’s
theorems in Part I, but I will also comment on some of the assumptions
that lead to Part II. There are alternatives to Penrose’s attempt to recon-
cile a neuroscientific account of consciousness with an antimechanistic
view that are simply not explored in SM.
In Part I, Penrose sets out four main positions on computation and
cognition (p. 12): (A) All thinking is computation; in particular, feelings
of conscious awareness are evoked merely by carrying out the appropriate
computations; (B) awareness is a feature of the brain’s physical action,
and whereas any physical action can be simulated computationally, com-
putational simulation cannot by itself evoke awareness; (C) appropriate
physical action of the brain evokes awareness, but this physical action
cannot be properly simulated computationally; (D) awareness cannot
be explained in physical, computational, or any other scientific terms.
Penrose does not want to abandon science, and so he rejects D. D must
involve some kind of antiscientific mysticism. He thinks that C is closest
to the truth. Not all physical action can be properly simulated computa-
tionally. Penrose argues that A must be rejected on the basis of Gödel’s
theorem.
The argument from Gödel’s theorem takes up much of Part I. The
form in which Penrose presents Gödel’s theorem bears a strong resem-
blance to a generalized form of the incompleteness theorem in Kleene’s
Introduction to Metamathematics (Kleene 1952, which is cited by Penrose at
various points in SM). Kleene’s theorem XIII (p. 302) uses a very general
notion of a formal system F, where the main condition on F is that its set of
‘provable formulas’ be effectively enumerable. Suppose that F contains
effectively given ‘formulas’ (q, n) which are supposed to express the
predicate P(q, n) which holds just in case Cq (n) does not halt. F is said to
be sound for P if whenever F proves (q, n) then P(q, n) holds. F is complete
for P if the converse is true. Kleene’s theorem (slightly weakened) is that
if F is a formal system which is sound for the predicate P, then it is not
complete for it. Specifically, there is a k such that Ck (k) does not halt but
F does not prove (k, k).
Penrose states his version of the theorem directly in terms of Turing
machine computations (pp. 74–75). For the Penrose-Turing version, sup-
pose that A is a Turing machine which is such that whenever A halts on
an input (q, n) then Cq (n) does not halt. Then for some k, Ck (k) does
not halt, but A does not halt on (k, k). As Penrose puts it, the soundness
of A implies the incompleteness of A. The proof of the theorem in SM
P1: ICD
052183782c10.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:14
is a variant of the standard diagonal argument for showing that the halt-
ing problem for Turing machines is not effectively decidable. Assuming
Church’s thesis, the Penrose-Turing theorem and the Kleene theorem
are interderivable.
There are a number of technical flaws in Penrose’s subsequent discus-
sion of the relation of the Penrose-Turing theorem to the usual presen-
tation of the incompleteness theorems. These flaws have already been
clearly described in some detail in Solomon Feferman’s review of SM
(Feferman 1996), and I will not discuss them here. It will be useful, how-
ever, to say a few things about the relationship in order to set the stage
for some of the critical remarks later.
In response to a key objection (Q10, pp. 95–97), Penrose notes the
special role that 1 sentences play in connection with Gödel’s theorems.
1 sentences are of the form (∀y)Ry, where R expresses an effectively de-
cidable (= recursive) property of the natural numbers and the intended
range of ‘y’ is the set of natural numbers. We can also consider 1 sen-
tences with free variables, for example, (∀x)Rxy. Gödel constructs a 1
sentence G(F) equivalent (in PA) to (∀x)¬ ProofF (x, g), where g is the
Gödel number of G(F). G(F) thus provably expresses of itself that it is
not provable in F. If h is the Gödel number of the formula 0 = 1, the
1 formula Con(F): = (∀x)¬ProofF (x, h) expresses that F is consistent.
Then Gödel’s first theorem says that if F is consistent, then G(F) is not
provable in F. It is possible to show that if F is sufficiently strong, then F
is consistent iff F is sound for 1 sentences. It follows that the hypothesis
in the first theorem is equivalent to the soundness of F for 1 sentences.
Under this hypothesis G(F) is true. We therefore obtain a version of the
first theorem in the form of the Penrose-Turing theorem: if F is sound for
1 sentences, then F is not complete for them. Gödel’s second theorem
says that if F is consistent, then F does not prove Con(F): that is, it does
not prove its own consistency. One can get from the first to the second
theorem by showing that F proves Con(F) ↔ G(F).
Penrose bases his antimechanist argument on the Penrose-Turing ver-
sion of the first incompleteness theorem. As does Penrose, I count myself
among those who think the incompleteness theorems are not irrelevant
to mechanistic views of the mind. The problem is to state precisely how
they are relevant. Penrose, as does Lucas, thinks they are relevant in the
following way: according to the Penrose-Turing theorem, A is incapable
of ascertaining that Ck (k) does not stop even though it does not. From
the knowledge of A and its soundness we can actually construct a com-
putation Ck (k) that we can see does not ever stop. Thus, A cannot be a
P1: ICD
052183782c10.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:14
When read in the context of other claims in SM, however, this response
is puzzling. Penrose professes in some passages to be a platonist about
mathematical truths and objects, but 1 truth does not need to be con-
strued as platonistic. A 1 truth predicate is definable in intuitionistic
arithmetic. In any case, the appeal to the truth of 1 sentences here does
not cohere with what Penrose says elsewhere. On p. 91 he says that if a
formal system is sound, then it is certainly -consistent. This is a differ-
ent notion of soundness from that discussed earlier, since -consistency
is stronger than consistency for 1 sentences. At this point he seems to
have in mind soundness for all arithmetical sentences, as one might ex-
pect in a platonistic antimechanist argument based on Gödel’s theorems.
In the discussion of Q18 (p. 112), Penrose allows that “we cannot properly
encapsulate ‘soundness’ or ‘truth’ within any formal system – as follows
from a famous theorem of Tarski.” It therefore appears that soundness is
to be taken as the truth of all sentences. Or is he thinking of all arithmeti-
cal sentences? By the time the reader reaches the discussion of platonism
toward the end of the book Penrose seems to be claiming that mathe-
matical truth is an absolute matter. He says (p. 418) that Gödel’s proof of
the incompleteness theorems argues “powerfully for the very existence of
the Platonic mathematical world. Mathematical truth is not determined
arbitrarily by the rules of some ‘man-made’ formal system, but has an ab-
solute nature, and lies beyond any such system of specifiable rules.” Thus,
what exactly is the response to Q10? And how should we understand G?
One might expect Penrose, as a platonist, consistently to emphasize
a global notion of truth in his antimechanist argument. Otherwise, why
not try to provide an antimechanist argument on intuitionistic or con-
structivist grounds? The problem is that it is never very clear how we
should understand Penrose’s platonism and its role in the antimechanist
argument. This may be a problem that Penrose inherits from the ap-
proach of Lucas, for platonism also played no role in Lucas’ antimechanist
argument.
It is instructive to compare the Penrose-Lucas argument briefly with
Gödel’s antimechanistic remarks. Gödel’s remarks are typically made in
connection with the second incompleteness theorem and with the ques-
tion of what will be needed for consistency proofs for mathematics. It is
odd that neither Lucas nor Penrose places much emphasis on the sec-
ond theorem in conjunction with the existence of consistency proofs
for PA and for other more extensive systems. Penrose does consider
(Q15, p. 107) what has become a well-known objection about the role
that consistency plays in antimechanist arguments derived from Gödel’s
P1: ICD
052183782c10.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:14
theorems. The objection, as he puts it, is this: the formal system F that
we choose to use might not actually be consistent. At least we might not
be certain that it is consistent. By what right can we therefore assert that
G(F) is ‘obviously’ true? A response to this objection, which neither Lucas
nor Penrose offers, is that we in fact do have (relative) consistency proofs
for PA and other systems. These results have been accepted into the body
of mathematics and they are holistically connected to it. It can be argued
that even if we do not have a consistency proof for any choice of F, the ex-
istence of the consistency proofs we do have constitutes evidence (which
is not to be equated with ‘certainty’) for the claim that we have insight
into objects, concepts, or truths that are abstract and nonfinitary relative
to formalism or mechanism. This appears to be one element of Gödel’s
view.
The argument is simply that it follows from the second theorem, ap-
plied to PA, that the concepts or truths needed for a consistency proof for
PA are not available in PA. One must step outside PA. To suppose that the
concepts or truths that can be adequately represented in PA are concrete,
finitary, immediately available to sensory intuition, and so on, is already
rather liberal. There is reason to believe that we should say this only for
some subsystem of PA. It follows that the concepts or truths involved in
the consistency proofs for PA will not possess these properties. As Gödel
states the argument, we will need ‘higher level’ concepts, that is, concepts
that are not about concrete objects in space-time, such as finite combina-
tions of symbols. We will need concepts that are about abstract ‘thought
structures’ or ‘meanings’. It will be necessary to go beyond reflection on
concrete signs and their properties and include reflection on meanings
or intentions. From this point of view, which could be either platonistic
or intuitionistic, the problem for mechanism is how we could be using a
mechanical procedure to grasp these abstract and nonfinitary concepts or
truths. This formulation should be compared with thesis G. I will refer to
this line of reasoning as ‘the argument from abstract objects’. It can be
elaborated in many directions.
The fact that every theorem-generating machine can be recast as a
formal system, and vice versa, helps to make this philosophical point
about mechanism clearer. In Hilbert’s original program, formal systems
were supposed to have a foundational advantage by virtue of the fact
that they involved only concrete, finitary objects in space-time that were
available to immediate perception. Our activities with them would be di-
rectly verifiable in sense perception, much as the activities of participants
in the Turing test would be directly verifiable in sense perception. This
P1: ICD
052183782c10.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:14
to predict what our best physical and biological theories will look like
in one hundred or two hundred years. One might emphasize just how
primitive our scientific understanding of consciousness is at this time.
Some have argued that it is comparable to Neanderthal man’s under-
standing of the theory of relativity. However that may be, it seems that
Penrose’s suggestions could not possibly help with his own alleged pla-
tonism. A conception of physics according to which the brain grasps the
abstract, acausal objects recognized to exist in platonism must indeed
be extremely sublime. In fact, are we not faced with a contradiction here?
For Penrose, we need a radically new conception of physics to see how
the brain could know about abstract objects. On the other hand, there
are philosophical views on which we require a radically new conception
of mathematics to see how this is possible. Both views seem to me to be
wrong.
Gödel sometimes appears to be attracted to a third possible position.
The mind, but not the brain, could have access to abstract objects that
are not located in the space-time causal order. Objects that can affect our
senses and that are located in the causal order are accessible to the brain.
The brain functions basically as a digital computer. Thus, the mind must
be distinct from the brain. This view suggests something like a Cartesian
or ‘substance’ dualism. It is a position fraught with many problems. Or
it may be that Gödel thinks there is a kind of Leibnizian preestablished
harmony between these different domains.
Yet another possibility can be found in dual-aspect theories of men-
tal phenomena. Earlier theories of this type are found in Spinoza and
Kant. Recent theorists include Donald Davidson, Thomas Nagel, and
P. F. Strawson. Davidson, for example, argues for anomalous monism,
which he distinguishes from nomological monism (e.g., materialism,
Penrose?), nomological dualism (e.g., Leibniz), and anomalous dualism
(e.g., Descartes). Recent dual-aspect theories often hold that conscious-
ness exhibits intentionality and that the language of intentionality or of
reason cannot be reduced to the language of neuroscience or physics,
even if there is only one kind of thing in the universe. Dual-aspect theorists
deny the existence of psychophysical laws. They claim that there are irre-
ducibly mental ways of grouping phenomena. One must presumably also
apply this claim to our mental activities in mathematics. A dual-aspect the-
ory might provide a way to maintain that some activities of mathematical
consciousness are noncomputational without having to invoke a subtle
physics that is supposed to explain how the brain could grasp abstract
objects.
P1: IWV
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:20
part iii
CONSTRUCTIVISM, FULFILLABLE INTENTIONS,
AND ORIGINS
225
P1: IWV
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:20
226
P1: IWV
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:20
11
§ 1 Introduction
In intuitionism there is a fertile confluence of ideas on mathematics,
meaning theory, and cognition. Philosophers interested in any one of
these latter areas of research would profit from studying intuitionism. In
this chapter I want to focus on several connections between intuitionism
and some recent, post-Wittgensteinian views in the philosophy of mind,
meaning, and language. The views I shall focus on are associated with the
claim that human cognition exhibits intentionality and with related ideas
in philosophical psychology. This tradition in the philosophy of mind,
meaning, and language differs significantly from the tradition in which
Michael Dummett has interpreted and expounded upon intuitionism. 1
Dummett has said that he is not attempting to portray accurately the views
of Brouwer and Heyting, and more than one commentator has noted that
Dummett’s view of intuitionism diverges widely in some respects from
‘traditional’ intuitionism. The manner in which it diverges seems to be
primarily in its view of human cognition and, specifically, in its view of
Work on this essay was supported by N.W.O. (Dutch National Science Foundation) research
grant 22–266. I thank Dirk van Dalen for sponsoring the grant and for many helpful dis-
cussions about intuitionism. I am grateful to Per Martin-Löf and Dag Prawitz for making
it possible for me to visit them at the University of Stockholm and for discussing with me
some of the ideas in the chapter. Thanks are also due to Anne Troelstra for discussing some
of the material and for making available some of Heyting’s unpublished correspondence.
It should not be assumed that any of these people would endorse the arguments expressed
in this essay. Finally, I thank an anonymous referee for History and Philosophy of Logic for
some helpful comments.
1 See especially Michael Dummett 1977, 1978, 1991, and 1993.
227
P1: IWV
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:20
2 Readers who are not well acquainted with traditional intuitionism will find central sources
of information in Brouwer 1975 and Heyting 1931 and 1971. An extensive survey of results
that contains some philosophical discussion is Troelstra and van Dalen 1988.
P1: IWV
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:20
and yet he does not understand the language. Such a view would make
meaning private and ineffable.
Dummett has also argued that meaning holism in the style of Quine
must be rejected if this argument is to lead to the revision of mathematics
and logic, because the question of justifying deductive practices cannot
arise for a meaning holist. Dummett has argued for a molecular meaning
theory.
Various aspects of these views have been widely discussed and criti-
cized. Any alternative to Dummett’s view should include responses to the
following problems. First, some critics (see especially Prawitz 1977) have
claimed that Dummett’s requirement of ‘full’ manifestability is too strong
and should be weakened. Another objection is that a finitist could apply
Dummett’s argument to intuitionism itself to show that its idealizations of
human practice, like those of the platonist, are unfounded (Wright 1982;
George 1988). In that case, Dummett’s arguments against the platonist
are not compelling. It has been argued that, in fact, Dummett’s argu-
ment for intuitionism is circular (George 1993). Troelstra and van Dalen
have suggested some problems that arise from the fact that Dummett’s
approach is focused too narrowly on logic and arithmetic (Troelstra and
van Dalen 1988, p. 851). There has also been some concern about how the
notion of intuition has disappeared from intuitionism as a consequence
of the focus on meaning-theoretic issues. It has been argued, on the basis
of the Kantian background of Brouwer’s work, that intuitionism needs a
nonsolipsistic notion of intuition (Parsons 1986). In a related vein, it has
been suggested that Dummett’s view omits or downplays the conscious,
experiential character of mental construction (Tieszen 1994a). More-
over, this concern about intuition does not have to preclude a theory of
meaning.
quite different from Brouwer’s view. Brouwer held not only that mental
constructions could not be adequately captured in formal systems, but
that they were fundamentally languageless.4 But if one claims that in-
tuitionistic constructions are not completely captured in our linguistic
behavior, it does not follow, as Brouwer sometimes suggests, that they are
altogether languageless. To claim that intuitionistic constructions either
are languageless or must be completely captured in our linguistic practice
is to pose a false dilemma. In what follows, I shall think of intuitionistic
constructions in terms of group intentionality, by which I mean the capac-
ity for groups of people to share the same intentions and to see objects or
states of affairs under the same meanings. Scientific theories are just sets
of propositions that are believed to be true by groups of people. Groups
of people come to see problems under the same meanings or contents
and pursue their research accordingly. I argue for a view like this in the
sections that follow.
Philosophers as different from one another as Quine, Searle, and
Jackson agree that observable linguistic behavior underdetermines what
we know, understand, or believe. It underdetermines the meaning of
expressions, as Quine has pointed out with his famous ‘gavagai’ exam-
ple. If we understand meaning in terms of constructions, then linguistic
behavior also underdetermines constructions. Once this is recognized,
philosophers will have different things to say about the matter. I think
the most defensible line to take in intuitionism is this: because observable
practice in using sentences underdetermines our knowledge or under-
standing and does not suffice to explain it, one must fill in the explanation
by making inferences to unobservable, inner processes or structures that
may be shared by different subjects. This need not involve introspection.
It is a pattern of reasoning that goes back at least as far as Kant and that is
now used widely in linguistics, cognitive science, and artificial intelligence
studies.
Alexander George, using some ideas from linguistics and philosophy
of mind, has already discussed some of the relevant points in relation
4 For a more detailed discussion of what this claim amounts to, see Dirk van Dalen 1991.
One of the interesting points made by van Dalen is that in Brouwer’s philosophy ‘causal
sequences’ can have higher or lower degrees of ‘egoicity’ or dependence on the indi-
vidual subject, ranging from lawless to lawlike sequences. Lawlike sequences are most
independent of the individual subject and are therefore most objective. Van Dalen sug-
gests that Dummett’s view can be associated with the part of intuitionism that concerns
logic and lawlike sequences. Lawless sequences, however, are highly egoic. My comments
in this chapter will be confined to what can be said about lawless sequences insofar as
there can be a science of these objects.
P1: IWV
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:20
5 For a useful general introduction to the notion of intentionality see Searle 1983. Most
writers on the subject, including Searle, do not consider the concept in the context of
mathematics. I discuss the concept in relation to constructive mathematics in Tieszen
1989 and in Chapters 12 and 13.
6 See Heyting 1931. Some of the Heyting-Becker correspondence is now published in
Troelstra 1990. I thank Professor Troelstra for making the unpublished Heyting-Becker
correspondence available to me.
P1: IWV
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:20
It should be noted that in the 1931 paper and in the correspondence with
Freudenthal, Heyting mentions the idea that a proposition can also be
thought of as a problem (as in Kolmogorov’s interpretation). A problem is
posed by an intention whose fulfillment is sought. Heyting suggests that
the intention-fulfillment distinction can be thought of in terms of the
problem-solution distinction, or the expectation-realization distinction
(see Chapter 13).
7 See, e.g., Searle 1980 and Jackson 1986; also, Tieszen 1997b, p. 254.
8 For a formulation of the natural deduction term calculus see Troelstra and van Dalen
1988, pp. 556–563.
P1: IWV
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:20
9 Husserl in particular distinguishes meaning as mere intention from the meaning associ-
ated with the fulfillment of the intention. See Husserl’s Logical Investigations, Investigation
I, §§ 14–15. This is also discussed in Chapter 12, § 5.
P1: IWV
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:20
proposition (or intention) is grasped and what is known when the truth
or falsity of the proposition is grasped.
The general claim that intentions determine extensions does not im-
ply that intentions always determine extensions in a manner appropriate
to constructivism. Dummett, however, seems to slip into such an under-
standing of intentions. He thinks that what one knows when one knows
the meaning of an expression (or intention expressed by a sentence)
is something that is like a method of verification or a construction pro-
cedure. Not only that, but in mathematics one knows just the kind of
construction procedure recognized by intuitionism. Consider how he
says the meaning theory underlying his argument is related to Frege’s
meaning theory (Dummett 1993, p. 128):
Frege’s argument is that the theory of reference does not fully display what it
is a speaker knows when he understands an expression – what proposition is
the object of his knowledge. I have here endorsed that argument, but have,
in addition, gone one step beyond it, by maintaining that, since the speaker’s
knowledge is for the most part implicit knowledge, the theory of sense has not
only to specify what the speaker knows, but also how his knowledge is manifested;
this ingredient in the argument here used for the necessity of a theory of sense
is not to be found in Frege.
This remark, when coupled with his claims about intuitionism and
full manifestability, shows how Dummett slips into thinking that inten-
tions must always be constructive means of or methods for arriving at
extensions.
The matter, however, is more complicated and depends on other fac-
tors.10 At a given stage of our knowledge, intentions in a particular do-
main may not be determinate enough to determine extensions construc-
tively. This still allows for the possibility that these intentions condition
our experience in the domain, even if they do so in a rather minimal way,
as is the case when there may be thinking or theorizing about a domain
but not much knowledge. If this is correct, then it should happen that
one sometimes has to find the appropriate construction procedure. It is
not always built in. Sometimes an appropriately determinate fulfillment
procedure is present at the outset. It is known from the beginning. Then,
in a sense, there need be no meaning that goes beyond the procedure.
We often start with a problem (intention), however, and then find the
fulfillment procedure later (if at all). It seems that for some intentions,
taken at face value, it will not be possible to find a fulfillment procedure,
that is, fulfillment for us humans. In other cases it is just not clear whether
constructions will be found or not. Situations of these latter types arise in
parts of modern mathematics, and especially in impredicative set theory.
But this does not rule out the meaningfulness of these parts of mathe-
matics, even if it does rule out constructive meaning. It only shows that
the meaning of these parts of mathematics is less adequate, determinate,
or clear.
Dummett argues that the problem with the extensional, truth-
conditional view of meaning, insofar as it is used to support platonism
or classical mathematics, is that it gives us a notion of meaning that is not
recognizable by us, or that transcends our knowledge or understanding.
I agree that there is a problem with purely extensional views of meaning.
The fact that there is a problem does not, however, entail Dummett’s
version of the view that meaning is determined by use. The alternative
I have presented defers to intensionalism and is thereby in sympathy
with Dummett’s comments about the extensional, truth-conditional view
of meaning, but it does not identify all intentions with constructions.
Dummett’s argument against classical mathematics does not go through
on the basis of his own meaning-theoretic preferences. If mathematical
propositions far beyond the pale of constructivism are nonetheless mean-
ingful, as I have argued, it is not necessarily because we know their truth
conditions. The meaning theory that makes it possible to say this is not an
extensional, truth-conditional meaning theory of the type that Dummett
criticizes. It is a meaning theory based on intentionality. On this view, our
grasp of the meaning of these propositions is less adequate, determinate,
or clear, but we can have some grasp of their meaning.
If knowledge of meaning is separated from knowledge of extensions
in the manner suggested, then saying that classical mathematics is inco-
herent, unintelligible, and illegitimate is too strong. Instead, a different
perspective on classical mathematics emerges. Classical mathematics is
not known to be contradictory, but it is also not constructive. One of
the central features of intentionality is just that statements can be about
objects without those objects’ necessarily having to exist. Classical math-
ematics makes meaningful statements about certain objects, but without
the same kind of direct evidence for the existence of the objects. There
are, however, ongoing research programs in classical mathematics. I do
not see how it can be claimed that these research programs are inco-
herent or meaningless. On the contrary, these programs must be de-
termined by certain intentions. Thus, it can still be said that intentions
in higher set theory determine how one can go on with the research
P1: IWV
0521837820agg.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:20
(i.e., with unfolding the meaning that is already given) even if they do
not determine this process through procedures that can be characterized
exclusively as constructivist.
Let it be admitted that intuitionism is engaged in its own idealizations
of human practice. It recognizes, for example, the existence of functions
that are not feasibly computable, of objects that are ‘abstract’ relative
to finitism, and of meaningful statements about such objects. The ide-
alizations of classical mathematics, especially of set theory, extend even
further. What we can say is that our grasp of the meaning of these propo-
sitions is even less adequate, less clear and distinct, relative to our own
finiteness and to our other limitations in understanding the abstract, the
transfinite, and the further reaches of what can be thought.
Suppose that constructive meaning is distinguished from other types of
meaning on the grounds that mathematical expressions can have mean-
ing apart from fulfillment. Then it is possible to count grammatically
well-formed expressions of unfulfilled mathematical intentions as mean-
ingful, along with all intentions that are more or less adequately fulfilled.
It is possible to recognize degrees or types of fulfillment, or degrees or
types of evidence. It is interesting that Heyting already recognized de-
grees of evidence in the context of intuitionism.11 He suggested that we
have the highest grade of evidence in the case of singular statements in-
volving small natural numbers. In the case of such statements as 1002 +
2 = 1004 we do not actually count, but we already use general reasoning
that shows that (n + 2) + 2 = n + 4. Lower grades of evidence accom-
pany the introduction of the notion of order type in connection with
the constructible ordinals, the intuitionistic notion of negation, the use
of quantification, the introduction of choice sequences, and the intro-
duction of species. In a manner reminiscent of Bernays (Bernays 1935),
I suggest that the evidence thins out even further in parts of mathemat-
ics that extend beyond traditional intuitionism. It is possible to allow for
different types of evidence and, in particular, to allow for forms of in-
direct evidence. At the same time, we should distinguish the claim that
intentions determine extensions from the claim that all intentions de-
termine extensions constructively, recognizing that the general notion
of construction is constrained by certain conditions but is not precisely
defined. By way of contrast, it seems that Dummett does not want to say
that what the BHK interpretation gives us is only the constructive mean-
ing of the logical constants. He does not seem prepared to distinguish
12
This chapter is based on a talk I gave at the conference Hermann Weyl: Mathematics, Physics
and Philosophy, held at UC-Berkeley in April 1999. I thank the conference organizer,
Paolo Mancosu, for inviting me to speak. For helpful comments I am indebted to Paolo
Mancosu, Solomon Feferman, Thomas Ryckman, John Corcoran, Martin Davis, and several
other members of the audience. I also thank Mark van Atten, Dagfinn Føllesdal, and an
anonymous referee for Philosophia Mathematica for their comments on a draft of this essay.
248
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
influence on Weyl. Weyl studied some of Husserl’s writings and the two
corresponded. Weyl’s wife, Helene, was a student of Husserl’s. Husserl’s
influence can be seen very clearly in a number of Weyl’s works. One
characteristic that makes Weyl so interesting is that he was able to take
philosophical ideas from thinkers such as Fichte and Husserl and develop
them in mathematical ways. I hope to show how the mathematical ideas
of Das Kontinuum and Weyl’s other works on the foundations of mathe-
matics were shaped and guided by his philosophical preferences. It seems
that Weyl’s general philosophical outlook did not derive principally from
Poincaré, Russell, Hilbert, or even Brouwer, although there are points of
contact with each of these thinkers. I do not discuss the more technical
side of Weyl’s work in any detail in this essay. It would in any case be dif-
ficult to surpass Feferman’s (Feferman 1988) excellent discussion of the
technical work in DK and van Dalen’s (van Dalen 1995) superb treatment
of Weyl’s technical variations on Brouwerian intuitionism.
I also hope to present Weyl’s ideas in a way that will make apparent their
relevance to recent issues in the philosophy and foundations of mathe-
matics. The ideas on intuition, vicious circularity, the epistemic approach
to the paradoxes, meaning, the continuum, and the intuitive-symbolic
distinction are of special interest. Weyl’s arguments on foundations are
derived from an important tradition in philosophy, and, in some respects,
they remain compelling.
mind dependent. They are mind independent in the sense that there are
aspects of them that we do not constitute or construct, that is, their sen-
sory component. They are mind dependent in the sense that their form is
due to us. Kant says he is a transcendental idealist and an empirical realist.
In sensory perception the mind is constrained in what it can form by the
‘matter’ of experience. Roughly speaking, Kant uses this form/matter
distinction to find a way between the traditional, naive forms of idealism
and realism that preceded him.
There are several additional features of Kant’s philosophy (see, e.g.,
Kant 1973) to which I shall refer at later points in the essay. One of
the most important features is Kant’s distinction between intuitions and
concepts. Kant says that concepts without intuitions are empty, and in-
tuitions without concepts are blind. Knowledge is a product of concepts
and intuitions. For Kant, intuition always refers to the faculty of receptivity:
intuition is sensory intuition. There are two basic forms of sensory intu-
ition, space and time. When Kant speaks of ‘pure intuition’ he has in
mind these two basic forms of intuition. Space is the form of outer in-
tuition and time is the form of inner intuition. Kant thought that the
basic form of outer intuition – space – was captured by (Euclidean) ge-
ometry. He linked the form of inner intuition – time – to the natural
number sequence and arithmetic; however, the exact nature of the link
is not clear in Kant’s writings. Concerning mathematics in particular,
Kant says that mathematical knowledge requires the “construction of
concepts in pure intuition.” This too is not as precise in Kant’s writings
as one might like it to be, but clearly Kant’s critique of pure reason, ap-
plied to mathematics, is meant to establish that where there can be no
construction of concepts in pure intuition there can be no mathematical
knowledge. Philosophers who work in the Kantian tradition generally try
to make determinations about where intuition (and hence knowledge)
ends and pure reason begins. These ideas are certainly part of the back-
ground of Weyl’s views on foundations, although they take on a more
specific form that can be traced to the influence of Fichte and especially
Husserl.
Weyl mentions Fichte’s work at several points in DK, in Philosophy of
Mathematics and Natural Science, and in other essays (see Weyl 1925, 1955).
Fichte was a post-Kantian transcendental idealist who, as did Kant, em-
phasized the role of the mind or of consciousness in the constitution of
knowledge. Fichte retained many of Kant’s basic ideas about epistemology
and developed them in various ways (see especially the Wissenschaftslehre
of 1794, contained in Fichte 1908–1912). For example, he agrees with
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
Kant that time and space are basic forms of intuition, but he is not as
closely wedded to the idea that the form of outer intuition is captured
in Euclidean geometry. This development would certainly have appealed
to Weyl, for Weyl wrote an entire book, Raum, Zeit, Materie (RZM), on
the “new amalgamation of time and space” required by relativity theory.
Fichte’s philosophy contains a detailed account of the structures and pro-
cesses of the transcendental ego that are needed to explain how there
could be knowledge of the ‘nonego’, that is, the ‘objective world’ with
its various general and specific features. As do many post-Kantian tran-
scendental idealists, Fichte rejects Kant’s notion of the noumenon or
‘thing-in-itself’. Fichte retains the objective world as phenomenon. In re-
taining the objective world Fichte seeks to avoid naive idealism, but he
wishes to explain the genesis of experience from the side of the self. He
recognizes that we have the belief that we are acted upon by things that
exist independently of us, but he argues that the ego must have the capac-
ity to produce this very belief. Fichte’s explanation of how the ego could
produce this belief is quite remarkable, and it contains far more detail
than we can go into here. There are many remarks in Weyl’s work that
reflect the general claim of transcendental idealism that we must start
with the transcendental ego and its acts and processes and determine in
a critical fashion what can be built up from this basis. We must determine
what kinds of structures and processes are involved in the production
of knowledge. Weyl contrasts this view with what he calls “naive absolute
realism.” Naive absolute realism is not concerned with the ego and its
acts or with what the ego can constitute or construct in its acts.
Similar themes in Husserl’s writing clearly have great resonance for
Weyl. Weyl studied with Husserl and throughout his career cited Husserl’s
work in his writings. In the Preface of DK he states flatly that he agrees
with Husserl’s views concerning the epistemological side of logic as they
are expressed in the second edition of Husserl’s Logical Investigations and
in Ideas I. In the Introduction to RZM (published in the same year as
DK), Weyl describes his general philosophical view in terms that are ba-
sically paraphrased from Husserl’s discussions of the phenomenological
reduction in Ideas I. He says, for example, that
we are only concerned in seeing clearly that the datum of consciousness is the
starting point at which we must place ourselves if we are to understand the ab-
solute meaning as well as the right to the supposition of reality. In the field of
logic we have an analogous case. A judgment which I pronounce affirms a cer-
tain set of circumstances; it takes them to be true. Here again the philosophical
question of the meaning of, and the justification for, this thesis of truth arises;
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
here again the idea of objective truth is not denied, but becomes a problem
which has to be grasped from what is absolutely given. “Pure consciousness” is
the seat of that which is philosophically a priori . . . what is immanent is absolute.
(Weyl 1918b)
1 Page numbers in citations of Weyl’s works refer to the English translations of his works
listed in the Bibliography.
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
It is not the purpose of this work to cover the “firm rock” on which the house of
analysis is founded with a fake wooden structure of formalism – a structure which
can fool the reader and ultimately, the author into believing that it is the true
foundation. Rather, I shall show that this house is to a large degree built on sand.
I believe I can replace this shifting structure with pillars of enduring strength.
They will not, however, support everything which today is generally considered
to be securely grounded. I give up the rest, since I see no other possibility. (Weyl
1918a, Preface)
Although this is primarily a mathematical treatise, I did not avoid philosophical
questions and did not attempt to dispose of them by means of that crude and
superficial amalgamation of empiricism and formalism which still enjoys consider-
able prestige among mathematicians. . . . Concerning the epistemological side of
logic, I agree with the conceptions which underlie Husserl. (Weyl 1918a, Preface)
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
the relation of number to space and time we may say that time, as the form
of pure consciousness, is an essential, not an accidental, presupposition
for the mental operations on which the sense of a numerical statement is
founded” (Weyl 1949, p. 36). Since Weyl starts with a philosophy accord-
ing to which the ordering of experience is imposed by the mind, it is not
surprising that he holds that the ordinal conception of number is basic
(see Weyl 1949, pp. 34–35).
This form of intuition of course has many instantiations in ordinary
sensory intuition. For example, I intuit a cup before me; then at a later
stage in time I have a successor intuition of it, and then a successor
intuition of that intuition. Another possibilitity is that I intuit a pencil;
then at a later stage in time I have a successor intuition of it, and then
a successor intuition of that intuition, and so on. Yet another possibility
is that I intuit a particular physical object and then a different physical
object in a later intuition, and then a different physical object in a still
later intuition, and so on. The point is that I can have an intuition, then
another intuition, then another, and so on. It does not matter what the
sensory intuition is of.
From the perspective of Husserl’s phenomenology we can make the
following observations about this conception of intuition: at later (succes-
sor) stages of intuition the earlier stages sink back in time but are retained
in an appropriately modified manner, indicating the importance of the
role of memory in our constructions. Intuition is finite. We humans (as
transcendental egos) carry out only finitely many acts of intuition. There
are no completed infinite sequences of intuitions or, if you like, no com-
pleted infinite sets of acts of intuition. Infinite sequences of intuitions do
not, as it were, have being. The notion of infinity for our intuitions can be
thought of only as potential, in the sense of becoming. In the simple case
under consideration it is clear that this will be a lawlike or rule-governed
becoming. It is determinate. The future course of experience is fixed: it is
simply the iteration of successor. In Husserl’s language, we could say that
the ‘horizon’ of possibilities of experience here is fixed. (Later we will
contrast this with Weyl’s reflections on choice sequences.) At any stage
in this form we obtain a definite or determinate object; for example, the
object expressed by (((|)s)s)s is clearly distinct from the object expressed
by ((|)s)s.
It is because this sequence is lawlike and gives definite objects, I think,
that Weyl says in DK that “the intuition of iteration assures us that the
concept ‘natural number’ is extensionally determinate” (Weyl 1919; Weyl 1987,
p. 110). The concept ‘natural number’ has a definite horizon. In fact, in
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
DK Weyl takes the natural numbers (but only the natural numbers) to be
a closed infinite totality and he accepts the associated principles of proof
and definition by induction for these objects. The ‘definiteness’ of the
extension of the concept is reflected in the fact that the principle of the
excluded middle is permitted in this case but is not permitted in the case
of other mathematical categories. In light of his later (after 1919) views on
lawlike and choice sequences, Weyl seems prepared to hold that even if
we cannot carry out an intuition that gives all the natural numbers, we can
still suppose that what is given by a lawlike sequence has ‘being’ whereas
what is given by a choice sequence is only ‘becoming’. He evidently thinks
that we can shift from the notion of ‘becoming’ to the notion of ‘being’
in the case of lawlike sequences precisely because we can appeal to the
fact that the future course of experience is determined by a law (see § 6).
The intention/fulfillment scheme is also at work here. There could
be a mere intention in which I am directed toward a particular number,
but I can also fulfill this intention by carrying out a process to obtain the
number. If the number is quite large, I can bypass its actual construction
by instead working with the law that generates it.
One of the most important features of this notion of intuition is not
explicitly noted by either Weyl or Husserl, but I think it is clearly quite
important for Weyl, especially in DK. It is that there is no vicious circularity in
intuition. Carrying out a sequence of intuitions step by step to obtain an
object (to obtain, for example, (((|)s)s)s, or an instantiation of this form
in ordinary sense perception) does not require or presuppose that we
construct some (infinite, completed) totality to which the object belongs.
It is, as it were, a principle of perception or intuition that this should not
be required. If it were necessary to refer to such a totality, then obtaining
the object would require obtaining the totality, but obtaining the totality
would require obtaining the object – a vicious circle.
In real analysis the least upper bound principle is one of the main
casualties of adhering to this view of intuition. This principle states that
every bounded set of real numbers has a least upper bound (l.u.b.). In
this case the object – the l.u.b. – is defined in terms of the totality to
which it belongs. One needs to know the set of upper bounds before one
can determine the least among them. The l.u.b. therefore presupposes
the set of all upper bounds of which it is itself an element. In DK Weyl
of course gives up the l.u.b. principle for sets of reals, but he sees that
one can often get by with the l.u.b. principle for sequences of reals (see
also Feferman 1988). In the case of sequences bounded above by a real
number the l.u.b. is definable by means of an existential quantifier that
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
ranges only over the natural numbers. The l.u.b. principle for sequences
thus holds at the first level of what Weyl calls the ‘mathematical process’
for forming sets and functions (see § 4) and Weyl is quite firm about not
proceeding beyond this level.
I think that Weyl’s critical view of impredicativity in mathematics, es-
pecially in DK, stems from this kind of view of knowledge and of what
the mind can construct. In genuine cases of knowledge, founded as they
are on intuition, we do not have vicious circles. There is no possibility of
knowledge of an object, of the fulfillability of intentions, where there is
a vicious circle. In contrast to the standard progression involved in ob-
taining knowledge in which an intention may terminate in an intuition
(partial or complete), there is no possibility of the experience of truth
where there are vicious circles. It is intuition that gives us a starting point
or foundation of knowledge, understanding, and meaning. One can see
why Weyl would be unhappy with the set-theoretic treatments of the nat-
ural numbers that were afoot. In such treatments one starts with a theory
of infinite, completed totalities that is riddled with vicious circularity and
then defines the natural numbers on the basis of these totalities. As does
Poincaré, Weyl thinks this is exactly the wrong way to proceed (see also
Chapter 14). The founding relation should be in the other direction.
In reading Weyl I have often wondered why he does not seem sensitive
to the fact that impredicative specifications in mathematics do not always
yield contradictions. On the grounds of their impredicativity, he sweeps
away the set-theoretic version of the l.u.b. principle in the same stroke as
the Russell paradox, the Grelling paradox, and others. The explanation, I
think, is that for Weyl it is intuition that matters and on this basis we should
avoid vicious circularity whether it is bound up with contradictions or not.
Where there is vicious circularity the claim to knowledge is an illusion.
Weyl’s view of vicious circularity shows his commitment to idealism and
the importance of constructive intuition. He automatically embeds the
vicious circle principle (“no totality can contain members definable only
in terms of itself”) in a theory of our cognitive acts and processes. ‘Defin-
ability’ is interpreted in terms of intuitive constructibility. The contrast
with Gödel’s realistic reading of the vicious circle principle, for example,
could not be more stark. Gödel says that “if . . . it is a question of objects
that exist independently of our constructions there is nothing in the least
absurd in the existence of totalities containing members which can be
described (i.e., uniquely characterized) only by reference to this totality”
(Gödel 1944). On such a realist view, Gödel argues, one can deny the
vicious circle principle.
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
that Weyl permits in DK are founded on the natural numbers and are
nonindependent. Weyl construes this in a precise manner in DK: the sets
and functions permitted are explicitly defined over the basic category,
that is, the natural numbers.
Weyl is rather strict in DK in following what he calls the ‘narrower pro-
cedure’ for forming sets and functions. Using the ‘mathematical process’
that he describes in Chp. 1, § 4 of DK, one can form from the basic cat-
egory a new object category consisting of the definable sets (sets of the
first level). If one permitted iteration of the mathematical process, one
could introduce sets of the second and higher levels. Weyl, however, will
not allow higher types in the manner of Russell. He is focused on anal-
ysis and says that analysis with level distinctions is artificial and unwork-
able. On the other hand, we could in effect ignore the level distinctions
by adopting Russell’s axiom of reducibility, but, not surprisingly, Weyl
finds the axiom of reducibility totally unacceptable. His conclusion is that
the only acceptable procedure is to restrict attention to definable sets of
the first level and simply accept the consequences (such as giving up the
set-theoretic l.u.b. principle for the real number system). He says that
the objects of the basic categories (natural numbers and rational num-
bers) remain “uninterruptedly the genuine objects of our investigation
only when we comply with the narrower procedure” (Weyl 1918a, Chp. 1,
footnote 24). Otherwise the profusion of derived properties and relations
would be just as much an object of our thought as the primitive objects.
In order to reach a decision about the judgments formed under the re-
striction of the narrower procedure we need only to survey these basic
objects. We would otherwise be required to survey all derived properties
and relations as well.
This is one point on which it seems to me that Weyl diverges from
Husserl. Husserl’s writings suggest that for mathematics in general we
could allow sets of different levels provided they were formed in the right
way. Husserl’s notion of constitution or construction would presumably
be wider than Weyl’s. It is interesting to note that correspondence be-
tween Weyl and Becker shows that they already disagreed about the scope
of the phenomenological conception of mathematical constitution and
knowledge (see Mancosu and Ryckman 2002).
We have already seen that Weyl says that some categories or concepts
are extensionally definite and some are not. The categories ‘natural num-
ber’ and ‘rational number’ are extensionally definite, for example, but
the category ‘property of a natural number’ is not. The category ‘real
number’, as it is understood in classical analysis, is not extensionally
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
the failure to recognize that the sense of a concept is logically prior to its exten-
sion is widespread today; even the foundations of contemporary set theory are
afflicted with this malady. It seems to spring from empiricism’s peculiar theory
of abstraction; for arguments against which, see the brief but striking remarks
in Fichte (1908–1912) and the more careful exposition in Husserl (Ideas I ). Of
course, whoever wishes to formalize logic, but not to gain insight into it – and
formalizing is indeed the disease to which a mathematician is most prey – will
profit neither from Husserl nor, certainly, from Fichte. (Weyl 1919; pp. 110–111
in Weyl 1987)
§ 5 Meaning in Mathematics
Weyl makes a number of pronouncements about what has meaning and
what does not have meaning in mathematics. These pronouncements can
be understood against the background of the ideas about intentionality
and intuition presented previously. We said that we are always directed
toward an object or state of affairs by way of a meaning or intention.
Let us call this meaning by virtue of which we are directed toward ob-
jects ‘meaningi ’. This designates the meaning as mere intention. If the
intention is fulfilled or fulfillable, we will say that we have ‘meaningf ’.
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
This is the kind of meaning associated with knowledge and genuine un-
derstanding, the meaning that is founded on intuition. This distinction
in made by Husserl in § 14, Investigation I, of the Logical Investiga-
tions. We should therefore also distinguish what is meaninglessi from
what is meaninglessf . Expressions containing mathematical terms may
be meaninglessf without being meaninglessi . Indeed, Weyl says that the
category ‘real number’ in classical analysis has a sense or meaning, which
gives modern mathematicians the illusion that they understand it. Weyl
must be saying that it has meaningi . This category, however, is not exten-
sionally definite. It lacks the kind of meaning that is founded on intuition.
A string of signs might also be meaninglessi . In the opening pages of
DK Weyl gives an example that seems to be of this type: the expression
“An ethical value is green.” The terms in this expression are incompatible
with one another because they are from the wrong semantic categories.
These kinds of issues about meaning and the (in)compatibility of cat-
egories are discussed at great length by Husserl in his work on pure
grammar and logic in the Logical Investigations. Since this expression is
meaninglessi there is no possibility of its being fulfilled in intuition: that
is, there is no possibility of its being meaningfulf . Husserl even distin-
guishes other grounds of meaninglessness that are prior to problems
about mixing expressions from incompatible semantic categories. These
latter forms of meaninglessness result from problems of pure formal syn-
tax. In his analysis of conditions for the possibility of the meaningfulness
and truth of expressions Husserl thus finds that there are conditions at
different levels, starting with conditions on pure formal syntax and the
compatibility of semantic categories, followed by conditions on consis-
tency, followed in turn by conditions on fulfillability.
It can be concluded from what has been said that expressions involv-
ing vicious circularity must be meaninglessf . They express intentions that
are in principle not fulfillable. I take it that this is what Weyl has in mind
in DK when he declares questions such as “Is ‘heterological’ heterolog-
ical or autological?” to be meaningless. This seems to be Weyl’s general
approach to the paradoxes. The paradoxes contain vicious circles, but
we do not find vicious circles in knowledge that is founded on intuition.
Hence, conceptions involving vicious circularity cannot have the kind of
meaning associated with knowledge or understanding. There is no possi-
bility of knowledge or of the experience of truth with such conceptions.
These matters about meaning, intuition, and paradox in Weyl and
Husserl are also discussed by da Silva (da Silva 1997), but it seems to
me that his account is not quite correct. He says, correctly, that Husserl
distinguishes meaning with respect to form from meaning with respect
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
t
…
figure 6
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
figure 7
…
…
figure 8
certainly, the intuitive and the mathematical continuum do not coincide; a deep
chasm is fixed between them. . . . So one might say that our construction of analysis
contains a theory of the continuum. (Weyl 1918a, p. 93)
Weyl says that in order to capture the fluid continuum we should replace
the element/set relation with the part/whole relation. There are in fact
many kinds of part/whole relations, and it happens that Husserl distin-
guished many of these and discussed their logic in Investigation III of
his Logical Investigations. The intuitive continuum would be what Weyl,
following Husserl, calls an ‘extensive whole’ (Weyl 1949, p. 52). An ex-
tensive whole is a whole whose parts are of the same lowest genus as the
undivided whole itself. If a particular temporal interval is an extensive
whole, then its parts must themselves be temporal intervals. This idea
conforms to our experience since we have no intuition of durationless
points. (Similar observations can be made about extensionless points in
space.) On the basis of this part/whole analysis we would not be able to
obtain real numbers as idealized, extensionless points, as purely concep-
tual objects that are not given in intuition.
Weyl sees that his own theory in DK will not account for the intuitive
continuum. He is also certainly sensitive to mathematical facts about the
real continuum. The rational numbers, for example, are dense on the real
line, and they may be all that is required for measurement in the physical
sciences, but it is a mathematical fact that the rationals do not exhaust
the reals. The arithmetically defined or lawlike sequences of DK will not
suffice for the real continuum. One needs arbitrary sequences. Weyl be-
lieves that Brouwer has a constructively acceptable way to obtain this
effect. Weyl thinks that through the introduction of choice sequences
Brouwer has provided a mathematical account of the continuum that at
least is closer to conforming to intuition. In Philosophy of Mathematics and
Natural Science he states it as follows:
How then do assertions arise which concern, not all natural, but all real numbers,
i.e., all values of a real variable? Brouwer shows that frequently statements of
this form in traditional analysis, when correctly interpreted, simply concern the
totality of natural numbers. In cases where they do not, the notion of sequence
changes its meaning: it no longer signifies a sequence determined by some law or
other, but rather one that is created step by step by free acts of choice, and thus remains
in statu nascendi. This ‘becoming’ selective sequence represents the continuum, or
the variable, while the sequence determined ad infinitum by a law represents the
individual real number falling into the continuum. The continuum no longer
appears, to use Leibniz’s language, as an aggregate of fixed elements but as a
medium of free ‘becoming’. (Weyl 1949, p. 52)
Similar remarks are made in many places in Weyl’s writings. For example:
A real number is not to be defined as a set, but as an infinite sequence of nested ra-
tional intervals whose lengths converge to 0. . . . The individual interval sequence,
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
determined in infinitum by a law, then produces the individual real number, while
the free choice sequence produces the continuum. (Weyl 1925, p. 134)
and other early writings Husserl distinguishes between the intuitive and
the merely symbolic and holds that most of our arithmetic knowledge,
along with a lot of other mathematical knowledge, is merely symbolic.)
Weyl says that “if mathematics is to remain a serious cultural concern,
then some sense must be attached to Hilbert’s game of formulae, and I
see only one possibility of attributing it . . . an independent intellectual
meaning.” In theoretical physics we have before us an example of a kind
of knowledge of a completely different character from the common or
phenomenal knowledge that expresses purely what is given in intuition.
In the case of phenomenal knowledge, every judgment has its own sense
that is completely realizable within intuition, but this is by no means the
case for the statements of theoretical physics. In theoretical physics it is
rather the system as a whole that is in question if confronted with experi-
ence. Weyl says that
theories permit consciousness to ‘jump over its own shadow’, to leave behind the
matter of the given, to represent the transcendent, yet, as is self-evident, only
in symbols. Theoretical creation is something different from intuitive insight; its
aim is no less problematic than that of artistic creation. Over idealism, which is
called to destroy the epistemologically absolute naive realism, rises a third realm,
which we see Fichte, for example, enter in the final epoch of his philosophizing.
Yet he still succumbs to the mystical error that, ultimately, we can nonetheless
apprehend this transcendent within the luminous circle of insight. But here, all
that remains for us is symbolic construction. (Weyl 1925, p. 140)
Weyl says that symbolic construction will never lead to a final result in the
manner of phenomenal knowledge. Phenomenal knowledge is subject to
human error but is nonetheless immutable by its very nature. Symbolic
construction, Weyl says, is not a reproduction of the given, but it is also
not the sort of arbitrary game in the void proposed by some of the more
extreme branches of modern art. Consistency is one of the principles that
govern symbolic construction in mathematics. Weyl says that
beside Brouwer’s way, one will also have to pursue that of Hilbert; for it is unde-
niable that there is a theoretical need, simply incomprehensible from the purely
phenomenal point of view, with a creative urge directed upon the symbolic repre-
sentation of the transcendent, which demands to be satisfied. (Weyl 1925, p. 140)
It may well always be that the sense of a clearly and unambiguously determined
object concept assigns to the objects of the nature expressed by the concept their
sphere of existence. But this does not make the concept an extensionally definite
one; that is, it does not ensure that it makes sense to consider the existing objects
that fall under the concept as an ideally closed, in itself determined and delim-
ited totality. This cannot be so if only because the wholly new idea of existence,
of being-there [Dasein], is added here, while the concept itself is only about a
nature, a being such-and-such [So-sein]. Seemingly the only reason why people
have been tempted to adopt the contrary assumption is the example of the ac-
tual thing in the sense of the real external world, which is believed to exist in
itself and to possess a composition determined in itself. If E is a property, clear
in its sense and unambiguously given, of objects falling under a concept B, then
“x has the property E” is a sentence that makes a claim about a certain state of
affairs concerning an arbitrary object x of the sort. This state of affairs does or
does not obtain; the judgment is in itself true or not true. . . . If the concept B is,
in particular, extensionally definite, then not only will the question “Does x pos-
sess the property E?” have a clear and unambiguous sense for an arbitrary object
x falling under B, but also the existential question “Is there an object falling
under B that possesses the property E”? Based on the generating process of the
natural numbers given to us in intuition, we hold on to the view that the concept
of natural numbers is extensionally definite, and likewise for rational numbers.
But concepts such as ‘object’, ‘property of natural numbers’, and the like are
certainly not extensionally definite. (Weyl 1921, pp. 88–89)
Note that Weyl allows that “x has the property E” is a sentence that makes
a claim about a certain state of affairs concerning an arbitrary object x of
the sort under consideration and that this state of affairs does or does not
obtain. The judgment is in itself true or not true. This suggests that even
where concepts are not extensionally definite we can obtain decisions
about them based on their ‘sense’ or meaning. Obtaining these deci-
sions would not be an arbitrary affair; nor would it be a matter of estab-
lishing formal consistency; nor would it simply be a matter of artistic cre-
ation. Its organ seems to be “seeing” in a wide sense, not just “creativity.”
It is just this kind of decidability of judgments, based on an intuition of
the concepts they contain, that the later Husserl has in mind in speak-
ing of the intuition of concepts. Weyl’s comments in this passage bear a
resemblance to Husserl’s ideas about meaning clarification and the pro-
cess of intuiting essences through free variation in imagination. If Weyl
P1: ICD
052183782c12.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:27
is correct, we still cannot derive any existence claims from this process
alone. Where does this leave us? I think it does leave us with a view that
is worth exploring, but there is no space to do so here.
§ 8 Conclusion
In Weyl’s own attempt to do justice to mathematics as a whole we seem
to be left with an irreconcilable split between the intuitive and the purely
conceptual, along with everything that is entailed by this split. Weyl thinks
that Brouwer has done justice to the intuitive continuum with the intro-
duction of choice sequences. His own arithmetic analysis in DK does not
do justice to the intuitive continuum, but Weyl was prescient to suggest
that it would suffice for applications in physics.
Weyl’s work in foundations presents us with an uncommonly deep anal-
ysis of mathematical cognition and provides a vivid illustration of the kind
of technical work that can result from taking idealism and constructive
intuition seriously.
P1: IWV
052183782c13.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:36
13
§1
A passage from some work of the logician Martin-Löf (1983–84, p. 231;
cf. Martin-Löf 1987, p. 417) does a good job of capturing the cognitive or
epistemological aspect of proof that I wish to discuss. Martin-Löf says that
the proof of a judgment is the evidence for it . . . thus proof is the same as
evidence . . . the proof of a judgment is the very act of grasping, comprehend-
ing, understanding or seeing it. Thus a proof is, not an object, but an act. This
is what Brouwer wanted to stress by saying that a proof is a mental construction,
because what is mental, or psychic, is precisely our acts . . . and the act is primarily
the act as it is being performed, only secondarily, and irrevocably, does it become
the act that has been performed. (Martin-Löf 1983–84, p. 231)
where we “bracket” the object because we do not assume that the object
of the act always exists. Phenomenologists are famous for suggesting that
we “bracket” the object, and that we then focus our attention on the
act (noesis) and act-content (or noema), where we think of an act as
directed toward a particular object by way of its content (or noema).
Whether the object exists or not depends on whether we have evidence
for its existence, and such evidence would be given in further acts carried
out through time.
We can capture what is essential (for our purposes) to the distinction
between act and content by considering the following cases: a mathe-
matician M might believe that , or know that , or remember that ,
where is some mathematical proposition. In these cases different types
of cognitive acts are involved – believing, knowing, remembering – but
they have the same content, expressed by . The act-character changes,
but the content is the same. Of course the content may also vary while the
act-character remains the same. The content itself can have a structure
that is quite complex. Also, when we say that the content is “expressed”
by we shall mean that the mathematical proposition is an expression of
the content of a particular cognitive act. Thus, there is a direct parallel
between intensionality, a feature of language or expression, and inten-
tionality, a feature of cognition, insofar as we are restricting attention to
those expressions which are expressions of cognitive acts. We should not
necessarily expect, for example, substitutivity salva veritae and existential
generalization to hold for inferences involving expressions of intentions.
We could start by regimenting our understanding of what forms can
take by focusing on the syntax of first-order theories. Thus when we say
that an act is directed toward an object we could express this with the usual
devices of first-order theories: individual constants and bound variables.
For example, a mathematician might believe that Sa for a particular S and
a particular a, or that (∃x)Sx for a particular S and a particular domain
of objects D. can also have the form (∀x)Sx, S ∨ T, and so on. The
restriction to first-order theories is not necessary. In fact, there are good
reasons for considering higher types of intentions and their fulfillments,
but I shall not discuss this matter in any detail here.
The idea that an act is directed toward an object by way of its content
has a direct analog in the thesis in intensional logic and mathematics that
intension determines extension. We should comment on this thesis now
in order to forestall some possible misunderstandings of the intentional-
ity of cognitive acts. In particular, when we say that intension “determines”
extension, or that an act is directed toward a particular object by way of
its content, should we take this to mean that the act-content provides a
P1: IWV
052183782c13.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:36
§2
We said that it is the function of a proof to provide evidence. But what
kind of evidence? Intentions may be fulfilled in different ways or to differ-
ent degrees. For knowledge in general we could consider the presence
or absence of the following types of evidence: a priori evidence, evidence
of “necessity,” clear and distinct evidence, intersubjective evidence, and
“adequate evidence.” In mathematics a fine-grained approach to ques-
tions about the adequacy of evidence is already found in proof theory,
where proofs are ranked in terms of computational complexity and other
measures. The classifications of proofs that result are surely relevant to
questions about the “processes” which produce M’s belief that . As we
are viewing the matter, if one insists on a “feasible” proof, a finitist proof,
P1: IWV
052183782c13.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:36
¬
·
·
∧ ¬
§3
The concept of proof described in (1) admits of substantial investigation
and elaboration. Among other things, the distinction between empty and
fulfilled intentions can be looked at in a number of ways. Fulfillments of
intentions, in Husserl’s philosophy, are also understood as realizations
of expectations (Tieszen 1989). One can already see the foundations
for this in ordinary perception. In ordinary perception I may have a
conception of some object without actually seeing it, so that the intention
directed to the object is empty. Even so, I could not help but have some
expectations about the object, given the set of background beliefs I would
have acquired up to that stage in time. The empty intention, that is,
can be viewed as a set of anticipations or expectations about the object
which are determined by background beliefs, memories, and so on, at
that stage in time. Having such expectation is a fundamental feature
of human cognition, especially in contexts where one is attempting to
acquire knowledge. Then the expectation(s) may either be realized or not
be. Attempting to gain knowledge about an object, that is, is like realizing
certain expectations about the object. Of course one’s expectations may
have to be corrected or refined in the growth of knowledge.
The same structure is present in mathematical experience. Thus, we
can think of mathematical propositions (which we are viewing as expres-
sions of intentions) as expressions of expectations. The fulfillment of the
intention is the realization of the expectations. Hence, another possible
way to think about an epistemological or cognitivist conception of proof
is to hold that
(2) A proof is a realization of a mathematical expectation.
Speaking of a proof as a realization of an expectation, that is, should
not be separated from the meaning that these terms have in a theory
P1: IWV
052183782c13.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:36
§4
It is possible to develop some of the foregoing ideas in detail in formal
theories of proofs or constructions. I do not mean to suggest, however,
that any particular formal theory will adequately capture the concept
P1: IWV
052183782c13.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:36
a ∈ N
0 ∈ N
a ∈ N
1010 is not obtainable by the rule even though it is an element of N. But
we know we can bring it to the form a for some a ∈ N. Or consider a
proposition such as 1010 + 52 = 52 + 1010 . We know how to obtain a
canonical proof. A noncanonical, and shorter, proof would be to show
that (∀x)(∀y)(x + y = y + x) by mathematical induction (which is an
elimination rule in Martin-Löf’s system) and then instantiate (which is
also an elimination rule).
Consider how one could justify the following conditional by using the
clauses (a)–(f): ((S ∧ T) → U) → (S→ (T→ U)). It might be helpful to
think of how the proof of the conditional would look in a proof tree or in
a Fitch-style system of natural deduction. Let a:((S ∧ T) → U). That is, a
is a proof that coverts any proof (b, c) of S ∧ T into a proof a((b, c)) of U.
We would like a method of fulfilling (realizing) the intention (expecta-
tion) (S→ (T→ U)), so let d:S and e:T. Define a construction k such that
k(d): T→ U. That is, (k(d))(e):U. We should set (k(d))(e) = a((d, e)), so
that, using the functional abstraction operator, k(d) = e.a((d, e)) and
P1: IWV
052183782c13.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:36
14
Poincaré is well known for claiming that logicist and formalist attempts to
provide a foundation for mathematics are misguided. He is also known
for holding that any mathematical results that depend on viciously circu-
lar impredicative definitions are illusory. In Poincaré’s work these views
are linked to a form of constructivism about the natural numbers, to the
idea that it is appropriate to reason about only those objects that can
be defined in a finite number of words, and to the related idea that in
order to be meaningful mathematical theorems must be finitely ‘verifi-
able’. Poincaré’s arguments on these matters have been influential, and
I believe we can still profit from thinking about them today. My own
viewpoint on the foundations of mathematics has been influenced by
Husserlian transcendental phenomenology, but I find that Poincaré and
Husserl have similar positions on a number of general epistemological
issues, and that some of their views reinforce one another.
In this chapter I shall discuss Poincaré’s views on logicism, impred-
icativity, and formalism from the perspective of some central themes of
Husserl’s transcendental phenomenology. Among the most important
themes in phenomenology are that mathematical cognition exhibits in-
tentionality, and that the meaning and reference of mathematical propo-
sitions must be understood accordingly. This is accompanied by the
view that there are various ‘meaning categories’ in our thinking, and
A shorter version of this chapter was presented as a plenary lecture at the International
Congress Henri Poincaré, Nancy, France, May 1992. I benefited from the helpful comments
of a number of the audience members, especially Moritz Epple, Gerhard Heinzmann, Jairo
da Silva, and Jan Wolenski.
294
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
§ 1 Logicism
Poincaré and Husserl were both critical of logicism, although they were
involved in the debate at different stages. Husserl’s target was Frege’s
Grundlagen der Arithmetik, which he subjected to criticism in the Philosophie
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
1 Page numbers in citations of Poincaré’s works refer to the English translations of his works
indicated in the Bibliography.
2 Given these remarks on the a priori nature of mathematical induction, and on how it
is distinct from empirical induction, it does not seem charitable to claim, as Goldfarb
(Goldfarb 1988) does, that Poincaré’s views are psychologistic.
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
criterion of analyticity in more strictly Kantian terms than Frege, that is,
in terms of identity or the principle of contradiction. Thus, he argues
(Poincaré 1908) that induction (‘the rule of reasoning by recurrence’)
is irreducible to the principle of contradiction. Analytic proof, which is
what Poincaré takes to be found in pure logic, is characterized as ‘sterile’
and ‘empty’, when contrasted with mathematics. Logic remains barren,
Poincaré says, unless it is fertilized by intuition.
Poincaré also criticizes logicist and Cantorian definitions of number
on the grounds that they are impredicative and subject to vicious circular-
ity. Although it is true that Frege’s system contained a contradiction and
that various paradoxes had appeared in set theory, I think that we must
now see the role of impredicativity in mathematics differently from the
way Poincaré saw it. The presence of impredicativity in mathematics, or
the failure to apply the vicious circle principle, does not necessitate para-
dox. We can, however, preserve something of Poincaré’s point even if we
agree that impredicativity itself does not necessitate paradox. It has been
argued that impredicative specification in the context of very abstract and
comprehensive systems of the sort put forward by logicists and set theo-
rists is incompatible with the idea that mathematical propositions must
be based on constructive intuition. If these arguments are correct and if
the natural numbers could be adequately defined in such comprehensive
systems, then one might argue, as Frege did, that propositions of arith-
metic do not depend on intuition. We can therefore see, from another
perspective, why Poincaré might criticize logicist and set-theoretic defini-
tions of the natural numbers on account of their impredicativity, even if
he overstated the case by declaring impredicative specifications in these
theories to be meaningless. I will return to the topic of impredicativity
later.
The phenomenological view also entails the denial of nearly every-
thing in our brief characterization of Frege’s position. Husserl explicitly
criticized Frege in the Philosophie der Arithmetik. I will discuss one part
of this early criticism by way of some of Husserl’s more mature views
in the Logical Investigations (second edition), Ideas I, Formal and Tran-
scendental Logic, and The Crisis of the European Sciences and Transcendental
Phenomenology.
A central feature of the phenomenological view is the claim that math-
ematical cognition exhibits intentionality. This means that mathematical
cognition is directed toward (or ‘referred’ to) mathematical objects or
states of affairs by way of the intentional contents (or ‘senses’) of its
acts, where the objects or states of affairs need not exist. This is, in some
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
they are intended. This makes his view more faithful to the actual practice
of mathematicians than the views of those who wish to define or derive
the principles of number from logic or set theory. If we take objects such
as natural numbers as they are intended in mathematical practice, then
we can see why Husserl claimed early on in Philosophie der Arithmetik that
it is a problem that the analyzed sense in Frege’s definition of number
is not the same as the original sense (see Tieszen 1990 and Chapter 15).
Number terms need not be taken as referring to Frege’s infinite equiva-
lence classes, nor to any other specific kinds of sets that can be used to
model the principles of arithmetic.
The phenomenological view generally implies that we should be cau-
tious about eliminative reductionist schemes in the foundations of math-
ematics. Reductionism is the effort to collapse the differences between
the meanings under which we think objects by introducing translation
schemes and the like. Problems can arise when reductionism is coupled
with eliminativist motivations. The logicist and set-theoretic efforts to
provide explicit, extensional definitions of numbers constitute one such
form of reductionism. In Husserl’s work there is a resistance to such
a one-sided reductionism and extensionalism about mathematics and
logic.3 The discovery of set-theoretic definitions of the natural numbers
is important, but there are infinitely many extensional definitions of the
natural numbers that satisfy the principles of arithmetic and do not pre-
serve extensional equivalence. A comment of Poincaré (1908, p. 154)
comes to mind here, even if it is somewhat overstated: “you give a subtle
definition of number, and then, once the definition has been given, you
think no more about it, because in reality it is not your definition that
has taught you what a number is, you knew it long before, and when you
come to write the word number farther on, you give it the same meaning
as anybody else.” Poincaré thinks the problem is compounded by the
fact that logicism and Cantorism start with notions that are unclear and
unfamiliar instead of starting with clear and familiar notions.
As we consider these ideas of phenomenology in relation to Poincaré,
we need to keep in mind several important points about meaning theory.
Both Husserl and Frege distinguish the meaning of an expression from its
reference, but there does not seem to be such a distinction in Poincaré’s
writing. Poincaré does not write extensively on meaning theory, but he
3 Husserl in fact favored the development of intensional logic, and in this he is perhaps
unlike Poincaré. Frege also came up against the idea of developing an intensional logic,
but he of course did not pursue the matter.
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
does say (1913, p. 62) that if a mathematical theorem is not finitely ver-
ifiable, then it is meaningless. ‘Verifiable’ evidently means that it must
be possible in principle to provide an instance for which the proposition
holds. General propositions for which we do not possess proofs, for in-
stance, Goldbach’s conjecture, would then be meaningful insofar as we
can at least verify the statement for an instance. Without attempting to
pursue Poincaré’s views on meaning any further, we can already draw
a contrast with the phenomenological view. On the phenomenological
view, propositions can have meaning quite independently of whether
or not we can find an instance for which the statement is true. Even if
we happen to be working under an illusion at a particular stage in our
mathematical experience, our experience at that stage is still meaningful
(as Frege discovered). It is clear from his Grundgesetze and from his re-
sponses to Husserl and others that Frege wanted to prevent intensionality
from obtruding in the science of mathematics itself. In matters of mean-
ing theory, however, Frege and Husserl both held that we should not
confuse lack of reference with lack of meaning, nor even logical incon-
sistency with lack of meaning. In the same vein, I think we must say that
meaningfulness could not be a function of the distinction between the
finite and the infinite, nor of the distinction between the predicative and
the impredicative. Poincaré’s view can probably not accommodate these
points. There appears to be no explicit mention of intentionality or of a
sense/reference distinction in Poincaré’s writing. This simple point about
Poincaré’s meaning theory has some important consequences which are
noted later.
For Poincaré, Husserl, and the post-logicist Frege, the view that arith-
metical propositions have an irreducible mathematical content or mean-
ing accompanies the claim that arithmetical knowledge depends upon an
a priori form of intuition. As the objections of Poincaré and Husserl sug-
gest, arithmetic truths are not known to be true by virtue of purely logical
(or analytic) relations between propositions alone. As Husserl would put
it, there must be intuitions of objects or states of affairs corresponding to
the meanings under which we think objects if we are to have arithmetic
knowledge. There must be some way of providing a referent correspond-
ing to the meaning. In Husserlian language, our arithmetic intentions
must be fulfillable. Husserl defines intuition in terms of the fulfillment
of our intentions. In order to support a knowledge claim about a partic-
ular object, there must be an intuition of that object. This bears a certain
similarity to Kant’s claim that all existence statements are synthetic. Ful-
fillable means fulfillable for us, so that the phenomenological view of the
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
§ 2 Impredicativity
Poincaré associated the paradoxes with the maladies he saw in logicism,
Cantorism, realism about sets, and an overreliance on a formal treatment
of sets. There could be nothing worse in science than to adopt princi-
ples that led to contradictions, and, as did Russell and Weyl, Poincaré
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
take the entities involved to be constructed by us. In that case there must
clearly exist a definition (i.e., the description of a construction) which
does not refer to a totality to which the object defined belongs, because
the construction of an object cannot be based on a totality of things to
which the object to be constructed itself belongs. If the construction of
an object were to be based on such a totality, we would simply never get
the object. To construct the object we would have to construct the col-
lection of which the object is a member, but constructing this collection
requires constructing the object. The circularity here is vicious.
Michael Hallett (Hallett 1984) has described this situation by saying
that there is no way to ‘complete’ an impredicative process, no matter
what powers of surveillance or of ‘running through in a finite time’ are
granted to a constructing agent. Folina (Folina 1992), explicating re-
marks of Poincaré, characterizes the matter by stating that in the case
of impredicative definitions that are viciously circular there is no way
to obtain a ‘determinate’ object. What it means for something to be an
object, however, is that it be determinate, that it have definite bound-
aries. Poincaré’s objection to impredicative definition is that it does not
give us definite objects, objects with determinate boundaries. Predicativ-
ity gives us determinate objects, and that is what constructivism requires.
Folina argues that impredicative definitions are therefore unfaithful to
our understanding of what an object is. I would argue that the situation
is more complicated than this, however, and that some additional distinc-
tions need to be taken into account. For example, one might distinguish
between what is determinate for us and what is determinate in itself.
In any case, since the VCP holds in the case of constructed entities, it
appears to provide one natural cutoff point for what should or should not
be admitted in constructive mathematics. Gödel argues, however, that if
we are realists about mathematical objects and meanings, then impred-
icative definitions by themselves are not a problem. Zermelo (Zermelo
1908) had similarly appealed to a form of realism to defend his axiomati-
zation of set theory against Poincaré’s charge that it allowed impredicative
specification of sets. As Gödel (Gödel 1944) puts it, “If . . . it is a question
of objects that exist independently of our constructions there is noth-
ing in the least absurd in the existence of totalities containing members
which can be described (i.e., uniquely characterized) only by reference
to this totality.” Gödel wants to recognize a difference between the def-
inition of an object, in the sense of a unique description of an object
already presumed to exist, and constructive definition, that is, the descrip-
tion of a process whereby we are to form a new object. Relative to such
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
a realist view, VCP can be denied. But there is reason, prima facie, to
think that we have here a limit on the idealized constructive intuition of
objects.
As I said, I shall not try to settle the question of whether the
predicative/impredicative distinction marks a limit on (idealized) con-
structive intuitability in set theory. Nor shall I address the matter of
whether Husserl is some kind of a predicativist concerning knowledge
of the particular objects of our mathematical intentions. Instead, I would
simply like to note that there are impredicative specifications in set the-
ory which are so far not known to lead to contradictions. A rich practice
in which there are many methods and results has developed in impred-
icative set theory. It seems to me that we are at least owed an explanation
of the conditions for the possibility of this practice.
If Poincaré’s thinking about impredicativity and intuition can be
faulted, it is not for failing to embrace a form of realism about the objects
of impredicative set theory in the style of Zermelo and Gödel, so much
as for ruling out the possibility of consistent, meaningful mathematical
thinking apart from constructive intuitability. Poincaré suggests, at times,
that mathematical propositions must either be founded on intuition or
be contradictory. In his criticism of Zermelo’s separation axiom in the “La
logique de l’infini,” however, he appears to back away from this strong
position. He notes (Poincaré 1913, p. 59) that Zermelo indeed does not
allow himself to consider the set of all the objects which satisfy a certain
condition because such a set is never ‘closed’. It will always be possible
to introduce new objects. Zermelo has no scruples, however, about ac-
cepting a set of objects which satisfy a certain condition and which are
part of a certain set s, for he believes that we cannot possess a set without
possessing at the same time all of its elements. Among these elements, we
choose those that satisfy the condition, without fear of being disturbed
by new and unforseen elements, since we are supposed already to have
these elements at our disposal. Poincaré says that Zermelo has erected an
enclosing wall with his separation axiom by positing the set s beforehand
to keep out intruders who would enter from without, but he objects that
Zermelo does not worry about whether there could be intruders from
within his wall (Poincaré 1913, p. 59–60) If s is infinite, then its elements
cannot be conceived of as existing beforehand, all at once. It is possible
for new elements to arise constantly, and they will simply arise inside the
wall instead of outside. Thus, although Zermelo has “closed his sheep-
fold carefully, it is not clear that he has not set the wolf to mind the
sheep.” What I would like to note is that Poincaré is not arguing here
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
§ 3 Formalism
I cannot treat everything that Poincaré and Husserl say about formalism,
and so I will focus on one general theme that emerges from their writings.
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
Instead of entering into issues about Poincaré’s effort to wield his petitio
argument against Hilbert’s formalism (see, e.g., Parsons 1965; Steiner
1975), I will focus on some of the problems alluded to earlier about the
role of meaning in mathematics. These problems are common to the
versions of formalism found in Thomae, Hilbert, Curry, and others.
In his discussion of Hilbert, Poincaré (Poincaré 1908, p. 147) notes that
for the formalist it is not necessary or even useful to know the meaning
of a theorem in order to demonstrate it. We might as well put axioms
into one end of a machine and take out theorems at the other end, a
procedure Poincaré disdainfully likens to a legendary sausage machine
in Chicago. He complains that, on this view, it is no more necessary for a
mathematician to know what he or she is doing than it is for a machine.
Poincaré (Poincaré 1908, p. 148; also p. 32, p. 51ff.) contrasts this kind of
formalism and mechanism with the case of “arguments in which our mind
remains active, in those in which intuition still plays a part, in the living
arguments, so to speak.” A theme that runs through many of Poincaré’s
papers is that on the basis of logic, formalism, or mechanism alone, it is
possible to combine mathematical ideas or entities in many different ways.
It is possible to create indefinitely many formal systems. The problem is
that most of these formal systems are useless and totally devoid of interest.
There must be a process of selection or discernment in order to obtain
the relatively small number of combinations that are useful. The nature
of mathematical discovery lies in this process of selection, and formalism,
mechanism, and logic alone cannot account for it.
We have already seen a suggestion about what the phenomenological
explanation of all of this would be. Husserl holds that where there is rea-
son there is thinking of objects under certain intentions or meanings.
These meanings govern our expectations about objects and, hence, our
investigations. Intentionality just is directedness. The horizons of possi-
bilities associated with mathematical acts are determined in a rulelike
way by these meanings. The meanings of our acts fix our expectations
and our understanding of how to work toward the ideal of completing
our knowledge, and it is on this basis that we set and (nonmechani-
cally) solve problems. Poincaré’s “living arguments in which the mind
remains active” are precisely the arguments in which this kind of mean-
ing plays a role. If this kind of meaning or intentionality were not in-
volved, there could be no selection or discovery. There would be no
directedness, no goal-oriented or purposive activity of the sort that we as-
sociate with reason. There would be no particular set of possibilities asso-
ciated with a meaning, as distinct from some other set of possibilities. The
P1: IWV
052183782c14.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:38
works such as the Krisis (Husserl 1936) is supposed to involve just such
a degeneration. On a strict formalist view, theoretical science amounts
to nothing more than the construction and mastery of formal systems.
Science becomes nothing but technique. The virtues required for its pur-
suit are narrow in character. They are the virtues of the technician who
acquires familiarity with formal systems. The internal goals of science for
such a technician shift from understanding and discovery to formal el-
egance and pragmatic success. The professional scientist is reduced to
a theoretical technician. Formalism, taken strictly, distances the sciences
from the informal problems and resources that are central to scientific
growth. Speaking of both formalism and logicism in La valeur de la science
(Poincaré 1905, p. 21), Poincaré says that “in becoming rigorous, math-
ematical science takes on a character so artificial as to strike everyone; it
forgets its historical origins; we see how the questions can be answered,
we no longer see how and why they are put.”
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
15
The work of Frege and Husserl on logic and mathematics might, from
a modern perspective, be compared under three main headings: math-
ematical logic, philosophical logic, and the foundations of mathematics.
Under the first heading there is little room for comparison. Frege sur-
passed nearly everyone in the history of logic in his technical achieve-
ments and discoveries. Husserl contributed virtually no technical work
to the development of mathematical logic. Under the second heading,
however, there is a great deal of room for comparison. Many of the is-
sues raised by Frege and Husserl involving language, meaning, refer-
ence, judgment, platonism about logic, and other matters are still actively
debated in research in philosophical logic.
Although there is some overlap among the three areas, the grounds
for comparison are different again in the foundations of mathematics.
Frege had a technical program for the foundations of mathematics in
his logicism and, as we know from the Grundlagen der Arithmetik, it was
a program shaped by certain philosophical ideas. The program was for-
mulated so precisely in the Grundgesetze der Arithmetik that it could be
seen to fail. That is, it was possible to derive a contradiction from the
basic laws of arithmetic as these had been formulated by Frege, and sub-
sequent attempts to repair the damage led to developments that were
further and further removed from Frege’s effort to derive the principles
of number from logic clearly and decisively. Very late in his career, but
still several years before Gödel established the incompleteness theorems
I would like to thank Bill Tait for some helpful critical comments on my views about Frege
and Husserl.
314
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
1 Page numbers in citations of Frege’s works refer to the English translations of his works
indicated in the Bibliography.
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
possibility that we will have overlooked something that might cast doubt
upon the proof of a proposition. Thus, we will finally be able to eliminate
the possibility of error in our reasoning, and to establish arithmetic on a
secure foundation.
The formal system from which Frege intended to derive the principles
of arithmetic is a classical higher-order logic that includes a theory of
extensions of concepts (or classes). Within this structure Frege wanted to
give an explicit definition of number. The intended interpretation of the
system of the Grundgesetze was by all indications to be extensional. Frege
had of course made the sense/reference distinction (Frege 1892) and
had used it to give an account of identity, but he was thereby also enabled
in the Grundgesetze to set aside the notion of sense, and to prevent inten-
sional aspects of meaning from obtruding. Numbers were to be defined
in terms of the extensions of concepts, but concepts in Frege’s philosophy
were themselves taken to be the references of concept-words. The differ-
ence between a concept and the extension of a concept for Frege does
not coincide with the distinction between sense and reference. Rather, it
amounts to a difference in saturation: concepts are not saturated and so
are not objects, but extensions of concepts are saturated and are objects.
Frege knew that dealing with oblique contexts would require an inten-
sional logic, but oblique contexts were not to occur in the Grundgesetze,
and Frege did not go on to develop an intensional logic. The argument
of the Grundlagen and his exchanges with Husserl also indicate that, on
the whole, he sided with extensionalist approaches to logic.
In any case, the definition of number would be an extensional defi-
nition and because the definiens would involve extensions of concepts
it would be reductive in nature, ‘reducing’ numbers to infinite equiva-
lence classes. Of course Frege thought that this reduction involved no
notions that were not purely logical; that he was appealing only to the
acknowledged ‘logical’ notion of extensions of concepts in his defini-
tion. Numbers would thus be logical objects. It is interesting, however,
that in the Introduction to the first volume of the Grundgesetze Frege had
expressed some concern that his basic law involving the notion of the
course-of-values of a function (the ill-fated Basic Law V), which includes
the notion of the extension of a concept as a special case, might not be
viewed as a law of logic. Basic Law V is of course the source of the contra-
diction in Frege’s system, given his other assumptions. In the Appendix
on the Russell paradox in the second volume of the Grundgesetze (Frege
1903), Frege says that he never concealed from himself the fact that Basic
Law V lacked the self-evidence possessed by the other basic laws.
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
As we know from the Grundlagen, Frege thought that many old and
fundamental philosophical questions about arithmetic were intimately
linked to his logicist program. The principal dispute the Grundlagen is
meant to settle, he says, is whether the concept of number is definable or
not (Frege 1884, p. 5). He argues that if there are independent grounds
for believing that the fundamental principles of arithmetic are analytic
(as opposed to synthetic), then these would also tell in favor of the prin-
ciples’ being provable and of number’s being definable. If his program
could be carried out, we would also see that true arithmetical proposi-
tions do not have some irreducible mathematical content, that they are
not synthetic a priori and do not depend on (pure) intuition as Kant had
held, that they are not empirical in nature as Mill had held, that various
theories of numbers as aggregates are mistaken, that efforts to identify
numbers with perceivable signs are mistaken, and so on. In addition,
Frege’s vehement attacks on psychologism are meant to show that num-
bers are not mental entities, that various efforts to understand numbers
by appealing to psychological processes such as abstraction are mistaken,
and that true arithmetical propositions cannot be construed as psycho-
logical laws. Frege thought it crucial to the development of the science
of logic to establish the objectivity of logic, and hence of number, against
the subjectivity of our ideas about number.
We should note that although in many of his writings Husserl uses the
term logic in a very wide sense, even going so far as to subsume mathe-
matics under logic in his general theory of deductive systems, we do not
find in his work anything like the picture that Frege presents. Husserl
often tends to think of logic in the tradition of Bolzano and others as
something like a general theory of science. Although Husserl speaks of
‘logic’ in this very wide sense, he could not of course be a Fregean logi-
cist because he tells us in later works such as Formale und transzendentale
Logik (FTL) that a phenomenological-constitutional foundation of formal
logic is needed, and that the philosophical basis of logic is to be found
in transcendental phenomenology. Moreover, in his own early work on
number in Philosophie der Arithmetik (PA) he had argued that the concept
of number cannot be reduced to logical notions in Frege’s sense.
I had already acquired the definite direction of regard to the formal and a first
understanding of its sense by my Philosophie der Arithmetik, which, in spite of its im-
maturity as a first book, presented an initial attempt to go back to the spontaneous
activities of collecting and counting, in which collections (“sums”, “sets”) and car-
dinal numbers are given in the manner characteristic of something that is being
generated originally, and thereby to gain clarity respecting the proper, the au-
thentic, sense of the concepts fundamental to the theory of sets and the theory of
cardinal numbers. It was therefore, in my later terminology, a phenomenological-
constitutional investigation; and at the same time it was the first investigation that
sought to make “categorial objectivities” of the first level and of higher levels (sets
and cardinals of a higher ordinal level) understandable on the basis of the “con-
stituting” intentional activities.
important insights that are not found in Frege. I shall return to the ideas
in this passage in § 4.
In PA Husserl argued that the concept of number was not definable,
that it was a primitive concept, and he criticized Frege’s construal of nat-
ural numbers as (infinite) equivalence classes. Husserl generally thought
of concepts as intensional entities, so what he means when he says that
the concept of number is not definable, in Frege’s terms, is that one
cannot define the ‘content’, or the sense (intension) of the concept of
number (Tieszen 1990). The sense of the concept of number is logically
simple or primitive, it cannot be reduced to any other sense, and so we
must investigate it by some other means. In PA Husserl raises, in effect,
an early version of the paradox of analysis for Frege’s view (Resnik 1980;
Tieszen 1990; Dummett 1991b). The problem, simply put, is this: logic,
for Frege, is supposed to consist of analytic propositions, is not supposed
to depend on intuition, and so on. If we are trying to define the sense of
an expression of number, however, then the sense of the definiens will
either be the same as or different from the sense of the definiendum. If
the sense is the same, the definiens simply repeats the definiendum so
that the definition is pointless. But if the sense of the definiens is differ-
ent from that of the definiendum, the definiens adds something to the
definiendum, in which case the definition is false. Thus, we cannot hope
to define the concept of number.
Husserl argued, on the other hand, that our understanding of the ex-
tension of the concept of number is not problematic since we apply the
concept with no difficulty. Since Frege characterizes only the extension of
the concept of number, his work falls short of the goal of a philosophical
analysis of the concept. Frege, however, argues that the concept of num-
ber is definable, that an explicit, extensional definition of the concept can
be given. Further, he argues that Husserl’s criticisms really apply to all of
the concepts of mathematics, but that these criticisms miss the point since
extensional definitions of concepts suffice for mathematical purposes.
The disagreement here reflects a more general division between Frege
and Husserl on issues involving intensionality and extensionality in logic
and mathematics, and to some extent on the different methods and pur-
poses of mathematics and philosophy. Husserl, from the beginning of his
career, sided with intensional logicians in taking logic to be concerned
with senses or meanings as such. Frege’s view, on the other hand, is suc-
cintly expressed in the following comment (Frege 1892–95, p. 122):
They [the Umfangs-logicians] are right when, because of their preference for the
extension of a concept to its intension, they admit that they regard the reference
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
of words, and not their meaning, to be essential for logic. The Inhalts-logicians
only remain too happily with the meaning, for what they call “Inhalt” (content),
if it is not quite the same as Vorstellung, is certainly the meaning (Sinn). They
do not consider the fact that in logic it is not a question of how thoughts come
from thoughts without regard to truth-value, that, more generally speaking, the
progress from meanings (Sinne) to reference (Bedeutung) must be made; that
the logical laws are first laws in the realm of references and only then mediately
relate to meaning (Sinn).
Logic for Frege is, in the first instance, concerned with truth. It is not
concerned with (Fregean) thoughts or senses as such, but with thoughts
insofar as they are true. As Mohanty has pointed out, Frege agrees with
Husserl that the extension of a concept presupposes the intension of
a concept, but he also takes the concept itself to be the reference of a
concept-word (Mohanty 1982). Concepts and truth-values are both refer-
ences for Frege. Husserl, on the other hand, favored the development of
intensional logic, and he took the intensionalist outlook to be especially
important for philosophy. Indeed, the concept of intentionality, which
is central to his philosophy, calls for an intensional logic, for the inten-
tionality of acts of believing, knowing, and so on, creates oblique con-
texts. In particular, an analysis of belief or knowledge involving number
is ipso facto an analysis of intensionality. The marks of intensionality in the
failure of substitution salva veritate and existential generalization to pre-
serve validity of inference are seen throughout Husserl’s later philosophy.
Frege, on the other hand, devoted little attention to the analysis of beliefs
or knowledge about number. He seemed unable, at least in his logicist
period, to grasp the possibility of a nonpsychologistic analysis of this type.
Thus, one senses that there were some deeper issues behind the Frege-
Husserl dispute and that, to some extent, each philosopher missed the
other’s point. Frege claims that he is not after the intension of the con-
cept of number anyway, so that Husserl misses the significance of his
project for mathematics. But Husserl is interested, in effect, in a logic of
meanings, or a logic of oblique contexts generated by knowledge and be-
lief about numbers, so that Frege misses the point about how important
the development of such an intensional approach might be for deeper
questions in the philosophy of arithmetic. Since Frege characterizes only
the extension of the concept of number, his work falls short of the goal
of a philosophical analysis of arithmetic.
It might be argued on Frege’s behalf that we do have definitions of
numbers in various standard set theories, so that something of Frege’s
original claim about the possibility of providing an extensional, reductive
definition of number remains, even if the reduction is not to ‘pure logic’
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
in Frege’s sense. I think this cannot be denied, and that this set-theoretic
reduction constitutes a very important development in the foundations
of mathematics. Even so, something of Husserl’s point that the sense of
the concept of number is logically simple or primitive, that it cannot be
reduced to any other sense, also remains. Consider, for example, the fact
that there are infinitely many nonequivalent extensional definitions of
natural numbers in set theory. A view such as Frege’s is saddled with the
problem that there are many reductions. The existence of nonequivalent
extensional definitions in mathematics is of course not confined to ele-
mentary number theory. To take a different example, real numbers may
be defined as Dedekind cuts, as the upper or lower members of such cuts,
as equivalence classes of convergent sequences or rational numbers, and
so on. A real number defined as a Dedekind cut, for example, is not iden-
tical with a real number defined as the lower member of a cut since the
former is an ordered pair whereas the latter is a member of such a pair.
Thus, no particular definition of this type suffices to capture the sense
of the concept of a real number. Each such definition is ‘reductive’ and
fails to include the other definitions that are supported by the meaning
of the concept. So we cannot suppose that we know what real numbers
are or know that we have gotten to the essence of real numbers on the
basis of any one of these definitions, even if some of the definitions do
seem more ‘natural’ than others.
Looking back on the Frege-Husserl dispute I think we can say that
the upshot of Husserl’s objection is that we must seek to understand the
meaning or intension of the concept of number in some other way, and
that it is philosophically important that we do so, even if various kinds
of explicit, reductive definitions can be given. I think that subsequent
work in the foundations of mathematics bears this out. The same point
can be made about other basic concepts of mathematics. This is of special
interest in recent times in the case of set theory since mathematics can be
‘reduced’ to set theory, but no one understands what having an explicit
definition of the concept of set would mean.
Frege was of course not persuaded by Husserl that we must try to
understand arithmetic and the meaning of arithmetic concepts in a dif-
ferent way. PA was just too clouded with psychologistic confusions to be
persuasive. What eventually did convince Frege was Russell’s paradox.
to Kant’s view that the source of arithmetical knowledge lies in a pure a pri-
ori form of intuition. There are of course some notable differences: Frege
separates the geometrical from the temporal source of knowledge only
to set aside the temporal source, whereas for Kant the temporal source
is linked to arithmetic, as distinct from geometry. Also, there might be
disagreement between Frege and Kant about the claim that a statement
of number contains an assertion about a concept, and about whether the
concept of number is definable and statements of number are provable.
There is also no trace of realism about mathematical objects in Kant’s phi-
losophy. As we shall see, Frege’s postlogicist position on the foundations
of arithmetic is also closer to Husserl’s views in some respects, although
here too there will be some notable differences.
mental objects. But they are also not ‘logical objects’ in Frege’s sense.
They are, however, ‘ideal’ or abstract objects, even though the objects of
the underlying, founding acts in which they have their origin may be ob-
jects of sense perception. Husserl has a more subtle analysis than Frege of
what abstract objects are, and of how we could be aware of them. In “The
Thought” (Frege 1918) Frege struggles to make sense of how we could
come to know about a particular kind of abstract, eternal, immutable
object – a thought – and his comments leave the matter shrouded in mys-
tery. Husserl, on the other hand, gives a phenomenological (and also a
transcendental) solution to this problem: abstract objects, and numbers
in particular, are to be understood as invariants in our mathematical expe-
rience, or in mathematical phenomena. Even if we have not clarified the
meaning of the concept of number completely, we can still say that num-
bers are identities through the many different kinds of acts and processes
carried out at different times and places and by different mathematicians,
and this is analogous to the fact that physical objects are identities in our
experience even though we do not see everything about them. Now some
phenomena simply do not sustain invariance over different times, places,
and persons. In the case of logical or sensory illusions, for example, what
we take to be present at one point in our experience is not sustained in
subsequent experience. Parts of our arithmetic experience, however, are
not at all like this, and they have in fact become quite stable across times,
places, and persons, just as other parts of our experience have stabilized.
So even if we have not brought the meaning of the concept of number
to full clarity, we can still say that numbers are ‘objects’ in the sense that
they are identities through the multiplicities of our own cognitive acts
and processes. If this were not so, the science of mathematics as we know
it would not be possible.
The sense of the ‘abstractness’ of numbers is derived from several
facts. Numbers could not be objects (identities) of sensation because
objects of sensation occur and change in space and time. They could
not be mental in nature because what is mental occurs and changes in
time. One could expand on these claims equally well by using either
Fregean or Husserlian arguments. Numbers are also identities that ‘tran-
scend’ consciousness in the sense that there are indefinitely many things
we do not know about them at a given time, on the analogy with our
knowledge of perceptual objects, but at the same time we can extend
our knowledge of them by solving open problems, devising new meth-
ods, and so on. They transcend conciousness in the same way that phys-
ical objects do. And, similarly, we cannot will them to be anything we
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
like; nor can we will anything to be true of them. They are mind inde-
pendent. On the object side of his analysis Husserl can therefore claim
to be a kind of realist about numbers. Numbers are not our own ideas.
At the same time he is also a kind of idealist on the subjective side of
his analysis because he has a constitutional account of our awareness and
knowledge of numbers and a critical perspective on classical metaphysical
realism.
Let us now focus on what the constitutional account of the awareness
and knowledge of numbers looks like. How is knowledge of these ideal
objects possible? To understand how the awareness and knowledge of
any kind of object is possible we must realize, Husserl argues, that various
forms of consciousness, such as believing and knowing, are intentional.
So numbers, as ideal objects, must be understood as the objects of acts
that are intentional. Intentional acts are directed to objects by way of
their contents or ‘noemata’. We can think of the contents associated
with acts as the meanings or senses under which we think of the objects.
So in the parlance of recent work on cognition and meaning, Husserl
wants to provide a theory of content (specific to arithmetic) in which the
origins of arithmetic content are taken to lie in more primitive, perceptual
‘founding’ acts and contents, where the idea is to determine the a priori
cognitive structures and processes that make arithmetic content possible.
So it is argued, for example, that ‘founded’ acts of abstraction from and
reflection on such underlying, founding acts and contents are an a priori
condition for the possibility of arithmetic content, and hence for the
awareness of number. To understand or to clarify the sense of the concept
of number, therefore, is not (or not only) to find an explicit, reductive
definition of number in some other mathematical theory. It is rather to
provide, among other things, a genetic account of the a priori conditions
for the possibility of arithmetic knowledge.
Husserl argues that the sense of the concept of number has its origin
in acts of collecting, counting, and comparing (i.e., placing objects into
one-to-one correspondence). Note that these are acts that are appropriate
to number and not, prima facie, to geometry. Insofar as we are aware of
numbers in these acts the acts must involve a kind of abstraction from and
reflection on our most primitive perceptual experiences with everyday
objects. They are ‘founded’ acts in the sense that they presuppose the
existence of more straightforward perceptual acts. The latter kinds of
acts could exist if there were no arithmetic, but the genesis of arithmetic
presupposes such straightforward acts (see Tieszen 1989). This is what
Husserl has in mind when, in the passage quoted previously, he speaks
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
the objective. And Frege may actually need a similar epistemological view
of abstraction to support his later account of number.
I also suggest, in response to views such as those expressed by Dummett
(Dummett 1991c), that Husserl’s remarks about numbers as aggregates of
featureless units in PA must be understood in the context of the effort to
provide a genetic analysis of the concept of number. That is, in speaking
of numbers as aggregates of featureless units Husserl is describing a stage,
perhaps even a fairly early stage, and one that is closer to sense perception,
in the genesis of our consciousness of numbers. The general project
here bears comparison to Quine’s description of the genesis of set theory
in The Roots of Reference. We need not assume that Husserl’s description
constitutes the final or highest stage in our consciousness of number,
especially in light of the clarification and development that has taken
place in number theory, mereology, and set theory since the time of
Husserl and Frege. I agree, however, that there are difficulties about
precisely how numbers are to be understood in PA, including those due to
obscurities surrounding the relationship between Husserlian ‘aggregates’
and Fregean ‘concepts’.
kind of subject matter (Bell 1990). The content stroke is used simply to
indicate what Frege later calls a thought, devoid of any assertoric force,
parallel with Husserl’s idea of the content or sense of an act. The shift
from judgment to content can be eludicated by way of Frege’s idea that
in intensional contexts a sentence no longer expresses a judgment or
possesses assertoric force, and no longer refers to any object or prop-
erty in the natural world; rather it refers to a sense. Its reference is now
the sense that it expressed before it was embedded in an intensional
context.
What little Frege has to say about the concept of knowledge is expli-
cated by way of the distinction between thoughts and judgments. As late
as 1924 (Frege 1924–25c) the view is expressed succinctly:
Of course Frege’s account of grasping a thought, and the role that such
grasping is supposed to play in securing reference or knowledge, differs
from Husserl’s view (see, e.g., Dummett 1991d). On Husserl’s view it is not
necessary that the thought itself be an object of an act of consciousness
in order for us to refer to either ordinary perceptual objects or to objects
such as numbers, although it is possible to reflect on the thought. The
path to an object does not require a detour through a different object of
consciousness.
Now a critical problem for Frege is this: how, and under what condi-
tions, is it possible to proceed, epistemically, from thought (or content)
to judgment in the case where the thought is about numbers, or other
mathematical objects? Frege says that a source of knowledge is what justi-
fies the recognition of a truth. On what grounds, however, can the recog-
nition of a truth about numbers be justified? Husserl’s views on founding
and founded acts and contents, and on the role of intuition in knowl-
edge, are meant to answer precisely these questions. Husserl defines the
concept of intuition in terms of fulfillment of (empty) act-contents and
argues that a condition for the possibility of knowledge is that there be in-
tuitions of objects at different levels in the hierarchy of acts, contents, and
intended objects. Husserl thus argues that there is intuition of abstract
or ‘ideal’ objects such as natural numbers, although this intuition will of
course be a form of founded intuition (Tieszen 1989). It is built up from
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
The law of the excluded middle, in its subjective aspect, . . . decrees not only that
if a judgment can be brought to an adequation . . . then it can be brought to either
a positive or a negative adequation, but . . . that every judgment necessarily admits of
being brought to an adequation – “necessarily” being understood with an ideality for
which, indeed, no responsible evidence has ever been sought.
These remarks have obvious implications for the way we should under-
stand our intuitive knowledge of the ‘necessity’ of certain statements of
logic, and they bear a striking resemblance to some constructivist views of
logic. In FTL and other works Husserl develops a ‘critique’, in a Kantian
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
§ 7 Conclusion
In his logicist period Frege portrayed genetic analysis as being fruitless for
mathematics and logic and as actually leading us away from mathematical
work. He repeated this charge in many contexts, applying it to Mill for
example in the Grundlagen and to Husserl in his review of PA. From the
preceding discussion, however, we see that what genetic analysis amounts
to in Husserl’s later work is quite different from what Mill had in mind,
and also from what Frege took it to be. It is, first of all, not a psychological
investigation. Rather, it is an investigation into the a priori conditions
for the possibility of the consciousness and knowledge of number. So
Frege would have to argue that genetic analysis, as we described it, has
no place in the philosophy of arithmetic. But on what grounds could
he make such an argument? It is true that genetic analysis might lead
us away from formal work, but this is because we cannot forget about
the intuitive and philosophical foundations of the concept of number,
and because we cannot exhaust the sense of the concept of number in
a particular formal system. There is a sense in which Frege’s ideal of
a strictly scientific method in arithmetic must be given up. We cannot
have a complete system of gapless formal proofs for numbers. Husserl’s
later view places additional emphasis on the informal, rigorous concept
analysis which is the source of formal work. Informal rigor and critical
analysis on the one hand, and the mathematician’s technical work on the
other, are complementary scientific activities. Thus, technical and formal
scientific work is not at all incompatible with the analysis of the origins of
concepts. It is just that such work is itself now seen to have its origins in
informal rigor, so that we cannot expect to supplant such informal rigor
with some formal system. On this basis technical work can be encouraged,
extensionalism need not be challenged, and so on.
A case can be made for the claim that the elements of a Husserlian
transcendental philosophy of arithmetic discussed are compatible with
the post-Fregean, post-Hilbertian, and post-Gödelian situation in the
foundations of mathematics. And they are compatible in a way that still
makes a kind of rationalism (or antiempiricism) about mathematics pos-
sible, thus preserving something of Frege’s own antiempiricism about
number. Moreover, they also preserve something of the realism or objec-
tivism about mathematics that Frege championed, but in a way that does
P1: KsF
052183782c15.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:42
Bibliography
337
P1: JtR
052183782bib.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:45
338 Bibliography
Deppert, W. (ed.), 1988. Exact Sciences and Their Philosophical Foundations. Frankfurt
am Main: Lang.
Detlefsen, M., 1990. “Brouwerian Intuitionism,” Mind 99, 501–534.
1992. “Poincaré Against the Logicians,” Synthese 90, 349–378.
Drummond, J., 1978–79. “On Seeing a Material Thing in Space: The Role of
Kinaesthesis in Visual Percpeption,” Philosophy and Phenomenological Research
40, 19–32.
1984. “The Perceptual Roots of Geometric Idealization,” Review of Metaphysics
37, 785–810.
Dummett, M., 1973. Frege: Philosophy of Language. New York: Harper & Row.
1977. Elements of Intuitionism. Oxford: Oxford University Press.
1978. “The Philosophical Basis of Intuitionistic Logic,” in M. Dummett, Truth
and Other Enigmas. Cambridge, Mass.: Harvard University Press, 186–201.
1981. The Interpretation of Frege’s Philosophy. Cambridge, Mass.: Harvard Univer-
sity Press.
1991. The Logical Basis of Metaphysics. Cambridge, Mass.: Harvard University
Press.
1991a. Frege and Other Philosophers. Oxford: Oxford University Press.
1991b. “Frege and the Paradox of Analysis,” in Dummett 1991a, 16–52.
1991c. Frege: Philosophy of Mathematics. Cambridge, Mass.: Harvard University
Press.
1991d. “Thought and Perception: The Views of Two Philosophical Innovators,”
in Dummett 1991a, 263–288.
1993. “What Is a Theory of Meaning? (II),” in M. Dummett, The Seas of Language.
Oxford: Oxford University Press, 34–93.
1994. Origins of Analytic Philosophy. Cambridge, Mass.: Harvard University Press.
Feferman, S., 1988. “Weyl Vindicated: Das Kontinuum 70 Years Later,” Atti del
Congresso Temi e prospettive della logica e della filosofia della scienza contemporanee,
Vol. I. Bologna, CLUEB, 59–93. Reprinted in Feferman 1998, 249–283.
1996. “Penrose’s Gödelian Argument,” Psyche 2, 21–32.
1998. In the Light of Logic. Oxford: Oxford University Press.
Feferman, S., et al. (eds.), 1986. Kurt Gödel: Collected Works, Vol. I. Oxford: Oxford
University Press.
1990. Kurt Gödel: Collected Works, Vol. II. Oxford: Oxford University Press.
1995. Kurt Gödel: Collected Works, Vol. III. Oxford: Oxford University Press.
Fichte, J. G., 1908–1912. Werke, 6 vols. Edited by F. Medicus. Leipzig: Meiner.
Field, H., 1989. Realism, Mathematics and Modality. Oxford: Blackwell.
Folina, J., 1992. Poincaré and the Philosophy of Mathematics. New York: St. Martin’s
Press.
Føllesdal, D., 1988. “Husserl on Evidence and Justification,” in R. Sokolowski
(ed.), Edmund Husserl and the Phenomenological Tradition. Washington, D.C.:
Catholic University of America Press, 107–129.
1995. “Gödel and Husserl,” in J. Hintikka (ed.), Essays on the Development of the
Foundations of Mathematics. Dordrecht: Kluwer, 427–446.
Frege, G., 1879. Begriffschrift, eine der arithmetischen nachgebildete Formelsprache des
reinen Denkens. Halle: L. Nebert. Translated by S. Bauer-Mengelberg in van
Heijenoort (ed.) 1967, 1–82.
P1: JtR
052183782bib.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:45
Bibliography 339
1884. Die Grundlagen der Arithmetik. Breslau: Koebner. Reprinted and translated
by J. Austin as The Foundations of Arithmetic, Evanston: Northwestern University
Press, 1978.
1892. “Uber Sinn und Bedeutung,” Zeitschrift für Philosophie und philosophische
Kritik 100, 25–50. Translated by M. Black in P. Geach and M. Black (eds.)
1970, 56–78.
1892–95. “Comments on Sense and Reference,” in Hermes et al. 1979, 118–125.
1893. Grundgesetze der Arithmetik, Vol. I. Jena: H. Pohle. Translated in part by
R. Furth as The Basic Laws of Arithmetic. Berkeley: University of California
Press, 1964.
1894. “Review of E. Husserl, Philosophie der Arithmetik,” Zeitschrift für Philosophie
und philosophische Kritik 103, 313–332. Translated by E. Kluge, in Mind 81,
1972, 321–337.
1903. Grundgesetze der Arithmetik, Vol. II. Jena: H. Pohle. Translated in part in
P. Geach and M. Black (eds.) 1970, 173–244.
1918. “Der Gedanke: Eine logische Untersuchung,” Beiträge zur Philosophie des
deutschen Idealismus 1, 58–77. Translated by A. Quinton and M. Quinton in
Mind 65, 1956, 289–311.
1924–25a. “A New Attempt at a Foundation for Arithmetic,” in Hermes et al.
1979, 278–281.
1924–25b. “Numbers and Arithmetic,” in Hermes et al. 1979, 275–277.
1924–25c. “Sources of Knowledge of Mathematics and the Mathematical Nat-
ural Sciences,” in Hermes et al. 1979, 267–274.
Friedman, M., 1999. Reconsidering Logical Positivism. Cambridge: Cambridge Uni-
versity Press.
2000. A Parting of the Ways: Carnap, Cassirer, and Heidegger. La Salle, Ill.: Open
Court.
Geach, P., and Black, M. (eds.), 1970. Translations from the Philosophical Writings of
Gottlob Frege. Oxford: Blackwell.
George, A., 1988. “The Conveyability of Intuitionism: An Essay on Mathematical
Cognition,” Journal of Philosophical Logic 17, 133–156.
1993. “How Not to Refute Realism,” Journal of Philosophy 90, 53–72.
Gödel, K., 1929. “On the Completeness of the Calculus of Logic,” in Feferman
et al. (eds.) 1986, 60–101.
1931. “On Formally Undecidable Propositions of Principia Mathematica and
Related Systems I,” in Feferman et al. (eds.) 1986, 145–195.
1933e. “On Intuitionistic Arithmetic and Number Theory,” in Feferman et al.
(eds.) 1986, 286–295.
1933f. “An Interpretation of the Intuitionistic Propositional Calculus,” in
Feferman et al. (eds.) 1986, 300–303.
*1933o. “The Present Situation in the Foundations of Mathematics,” in
Feferman et al. (eds.) 1995, 45–53.
1934. “On Undecidable Propositions of Formal Mathematical Systems,” in
Feferman et al. (eds.) 1986, 346–371.
*1938a. “Lecture at Zilsel’s,” in Feferman et al. (eds.) 1995, 87–113.
*193?. “Undecidable Diophantine Propositions,” in Feferman et al. (eds.) 1995,
164–175.
P1: JtR
052183782bib.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:45
340 Bibliography
1940. “The Consistency of the Axiom of Choice and of the Generalized Con-
tinuum Hypothesis with the Axioms of Set Theory,” in Feferman et al. (eds.)
1990, 33–101.
*1941. “In What Sense Is Intuitionistic Logic Constructive?” in Feferman et al.
(eds.) 1995, 189–2000.
1944. “Russell’s Mathematical Logic,” in Feferman et al. (eds.) 1990, 119–
143.
1946. “Remarks Before the Princeton Bicentennial Conference on Problems
in Mathematics,” in Feferman et al. (eds.) 1990, 150–153.
1947. “What Is Cantor’s Continuum Problem?” in Feferman et al. (eds.) 1990,
176–187.
1949. “An Example of a New Type of Cosmological Solutions of Einstein’s Field
Equations of Gravitation,” in Feferman et al. (eds.) 1990, 190–198.
1949a. “A Remark About the Relationship Between Relativity Theory and Ide-
alistic Philosophy,” in Feferman et al. (eds.) 1990, 202–207.
*1951. “Some Basic Theorems on the Foundations of Mathematics and Their
Implications,” in Feferman et al. (eds.) 1995, 304–323.
*1953/59, III and V. “Is Mathematics Syntax of Language?” in Feferman et al.
(eds.) 1995, 334–363.
1958. “On an Extension of Finitary Mathematics Which Has Not Yet Been
Used,” in Feferman et al. (eds.) 1990, 240–251.
*1961/? “The Modern Development of the Foundations of Mathematics in the
Light of Philosophy,” in Feferman et al. (eds.) 1995, 374–387.
1964. “What Is Cantor’s Continuum Problem?” in Feferman et al. (eds.) 1990,
254–270. Revised version of Gödel 1947.
1972. “On an Extension of Finitary Mathematics Which Has Not Yet Been
Used,” in Feferman et al. (eds.) 1990, 271–280. Revised version of Gödel
1958.
1972a. “Some Remarks on the Undecidability Results,” in Feferman et al. (eds.)
1990, 305–306.
Goldfarb, W., 1988. “Poincaré Against the Logicists”, in P. Kitcher and W. Aspray
(eds.), History and Philosophy of Modern Mathematics, Minneapolis: University
of Minnesota Press, 61–81.
Greenberg, M. J., 1993. Euclidean and Non-Euclidean Geometries, 3rd ed. New York:
W. H. Freeman.
Gurwitsch, A., 1974. Phenomenology and the Theory of Science. Evanston, Ill.: North-
western University Press.
Hallett, M., 1984. Cantorian Set Theory and Limitation of Size. Oxford: Oxford Uni-
versity Press.
Hardy, L., and Embree, L. (eds.), 1993. Phenomenology of Natural Science. Dordrecht:
Kluwer.
Hauser, K., forthcoming. “Gödel’s Program Revisted,” in Husserl Studies.
Heelan, P., 1983. Space Perception and the Philosophy of Science. Berkeley: University
of California Press.
1989. “Husserl’s Philosophy of Science,” in J. N. Mohanty and W. McKenna
(eds.), Husserl’s Phenomenology: A Textbook. Lanham, Md.: Center for Advanced
Research in Phenomenology/University Press of America.
P1: JtR
052183782bib.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:45
Bibliography 341
Heidegger, M., 1962. Being and Time. Translated by J. Macquarrie and E. Robinson.
New York: Harper & Row. Originally published in German in 1927.
Heinzmann, G., 1997. “Umfangslogik, Inhaltslogik, Theorematic Reasoning,” in
E. Agazzi and G. Darvas (eds.), Philosophy of Mathematics Today. Dordrecht:
Kluwer, 353–361.
Hermes, H., Kambartel, F., and Kaulbach, F. (eds.), 1979. Gottlob Frege: Posthumous
Writings. Translated by P. Long and R. White. Chicago: University of Chicago
Press.
Heyting, A., 1931. “Die intuitionistische Grundlegung der Mathematik,” Erkennt-
nis 2, 106–115. Translated by E. Putnam and G. Massey in Benacerraf and
Putnam (eds.) 1983, 52–61.
1962. “After Thirty Years,” in E. Nagel, P. Suppes, and A. Tarski (eds.), Logic,
Methodology and Philosophy of Science. Stanford, Calif.: Stanford University
Press, 194–197.
1971. Intuitionism: An Introduction, 3rd ed. Amsterdam: North-Holland.
Hilbert, D., 1926. “Über das Unendliche,” Mathematische Annalen 95, 161–190.
Translated by S. Bauer-Mengelberg in van Heijenoort (ed.) 1967, 367–
392.
Hill, C., and Rosado-Haddock, G., 2000. Husserl or Frege? La Salle, Ill.: Open Court.
Husserl, E., 1891. Philosophie der Arithmetik. Psychologische und logische Untersuchun-
gen. Halle: Pfeffer.
1900. Logische Untersuchungen, Erster Theil: Prolegomena zur reinen Logik.
Halle: M. Niemeyer. 2nd ed. 1913.
1901. Logische Untersuchungen, Zweiter Theil: Untersuchungen zur Phänomenologie
und Theorie der Erkenntnis. Halle: M. Niemeyer. 2nd ed. in two parts 1913/
22.
1911. “Philosophie als strenge Wissenschaft,” Logos I, 289–341.
1913. “Ideen zu einer reinen Phänomenologie und phänomenologischen
Philosophie,” Jahrbuch für Philosophie und phänomenologische Forschung 1, 1–
323. Published simultaneously as Separatum by M. Niemeyer.
1928. “Vorlesungen zur Phänomenologie des inneren Zeitbewusstseins,” edited
by M. Heidegger in Jahrbuch für Philosophie und phänomenologische Forschung 9,
viii–x, 367–498. Published simultaneously by M. Niemeyer.
1929. “Formale und transzendentale Logik. Versuch einer Kritik der logischen
Vernunft,” Jahrbuch für Philosophie und phänomenologische Forschung 10, v–viii,
1–298. Published simultaneously as Separatum.
1931. Méditations cartésiennes. Introduction à la phenoménologie. Paris: Librairie
Armand Colin.
1936. “Die Krisis der europäischen Wissenschaften und die transzendentale
Phänomenologie,” Philosophia 1, 77–176.
1939. Erfahrung und Urteil: Untersuchungen zur Genealogie der Logik. Edited by
L. Landgrebe. Prague: Akademia Verlagsbuchhandlung.
1950a. Cartesianische Meditationen und Pariser Vorträge. Husserliana I. 2nd ed.
1973. The Hague: Nijhoff.
1950b. Ideen zu einer reinen Phänomenologie und phänomenologischen Philosophie.
Erstes Buch: Allgemeine Einführung in die reine Phänomenologie. Husserliana III.
Revised ed. in two parts 1976. The Hague: Nijhoff.
P1: JtR
052183782bib.xml CY530B/Tieszen 0 521 83782 0 March 11, 2005 13:45
342 Bibliography
Bibliography 343
344 Bibliography
Bibliography 345
346 Bibliography
Bibliography 347
348 Bibliography
Index
349
P1: IWV
052183782ind.xml CY530B/Tieszen 0 521 83782 0 November 3, 2005 4:32
350 Index
Index 351
352 Index
Index 353
informal rigor, 49–50, 66, 135, 143, 198, Kant, I., 25, 57, 62, 65, 88, 95, 101, 107,
298, 334 122, 145, 224, 228, 233, 235, 246, 248,
instrumentalism, 42 249–250, 253, 283, 284, 297, 300, 305,
intensions 318, 325
see concepts; Frege, G., meaning, theory on intuition and concepts, 250, 281
of; meanings; Weyl, H. Kaufmann, F., 8
intentionality, 9, 37, 51–54, 95–97, 126, Kim, J., 59
147, 150, 155–157, 161–162, 171–172, Kleene, S., 216, 287, 289
174, 180, 188, 189, 199, 211, 221, 223, Klein, F., 70, 77
224, 227, 237, 241, 242, 252, 261, Erlanger program, 76
279–293, 294, 297, 303, 311, 321, knowledge
327 causal theory of, 58
derived, 240, 288 of mathematical objects, 58–65
group, 235, 236 and proof, 276–293
intentional stance, 171 see also evidence; intuition
intentions Kolmogorov, A. N., 239
and intuitionistic mathematics, 237–247 interpretation of intuitionistic logic, 287
see also intentionality Kreisel, G., 49, 93, 97, 101, 109, 113, 334
intersubjectivity, see solipsism Kripke models, 289, 291–293
intuition, 29, 32, 61–62, 96, 98, 102, 252,
300 least upper bound principle, 257, 263
analogies between sensory and Leibniz, G., 24, 25, 80, 89, 93, 95, 129, 154
categorial, 75 Leibnizian preestablished harmony, 224
in arithmetic, 329–335 Lie, S., 70
categorial, 3, 128, 138 lifeworld (Lebenswelt), 35–36, 38, 85, 305
as distinct from conjecture, 75 lived body, 37, 83
of essences, 12, 69–89, 98, 128, 129, logic
273 intensional, 3, 162, 242, 280, 299, 317,
and impredicativity, 306–308 320–321
in intuitionism, 228, 233, 246 of noncontradiction in Husserl, 27–29
mathematical, 7, 63–64 as science of all possible sciences, 24–25,
of particular mathematical objects, 61 318
and vicious circularity, 257–258 as a theory of reason, 24–26
of sets, 76 threefold stratification of in Husserl,
see also evidence; Frege, G.; Gödel, K.; 26–30, 82
Maddy, P.; Poincaré, H.; Weyl, H. transcendental, 35
intuitionism, 62, 146, 182, 227–247 of truth in Husserl, 29–30, 332
and Husserl, 7–8 logicism, 230, 295–305
in relation to classical mathematics, see also Frege, G.; Poincaré, H.
244–246 Lotze, H., 24
see also Brouwer, L. E. J.; constructivism; Lucas, J. R., 215, 217, 220, 221
Dummett, M.; Heyting, A.
intuitive/conceptual distinction, 7 Maddy, P., 59, 108
see also Kant, I. and the analogy between perception of
intuitive/symbolic distinction, 3, 33, 254, sets and perception of physical
271–275 objects, 206–208
see also authentic/inauthentic on Benacerraf, 204, 208, 209
distinction on compromise platonism, 201–214
invariants, 55–58, 71–75, 76–83, 88–89, monistic and dualistic, 211–213
166, 169, 170, 171–172, 206, 326 on Gödelian Platonism, 201–202, 203
levels of, 74 on intrinsic and extrinsic justification,
202, 211
Jackson, F., 235, 241 on the justification of set-theoretic
judgments, 26–30, 34 axioms, 209–211
P1: IWV
052183782ind.xml CY530B/Tieszen 0 521 83782 0 November 3, 2005 4:32
354 Index
Index 355
356 Index
Index 357